Cryptopolitan

Square Verified+
Crypto news that doesn't waste your time. Breaking updates, market analysis, on-chain insights. Building the smartest crypto community.
Creator of the Year
Creator of the Year
1 Following
161.6K+ Followers
573.6K+ Liked
1 Badges
Posts
PINNED
·
--
At Cryptopolitan, we research, analyze, and deliver news—daily. From breaking updates to in-depth analysis, educational guides, and market insights, we’re here to keep you informed with neutral and authentic news. Thank you for trusting us to be your go-to source!
At Cryptopolitan, we research, analyze, and deliver news—daily.

From breaking updates to in-depth analysis, educational guides, and market insights, we’re here to keep you informed with neutral and authentic news.

Thank you for trusting us to be your go-to source!
Seoul to search AI chat logs after exam to catch cheatersSouth Korea intends to look through AI chatbot records after its upcoming college entrance exam to find students who used AI to get answers. The exam is on Thursday and the plan so far depends on an unfinished deal with four tech firms and on a law that has not even been written yet. What is South Korea planning to do about AI cheating on its college exam? Every year, South Korean students take a national college entrance test called the Suneung, also known as the College Scholastic Ability Test (CSAT). President Lee Jae Myung says the government will check AI chatbot records after this year’s test to find students who took photos of exam questions and asked AI for the answers.  He shared the plan on X and at an emergency video meeting he led from Manila, where he was attending a summit of Southeast Asian nations. When a student sends a photo of a question to a chatbot, the image goes to the company’s servers and leaves a record of use. With this new rule, if investigators suspect a student of cheating, they can ask the company for that record.  Science Minister and Deputy Prime Minister Bae Kyung-hoon said four global tech companies have agreed to help when the South Korean government asks, but he admitted the companies have not agreed on how far that help will go.  The plan would require a new law capable of forcing AI companies to hand over records of illegal requests for exam answers. Lee has ordered that law, but it has not been passed. The Korean government considered blocking the chatbots during the test but dropped the idea. China’s main AI services briefly refused to solve photos of exam questions during its college entrance exam in June 2025.  Deputy Prime Minister Bae said Korea cannot easily copy China’s approach, because it would require foreign AI firms to build a filter just for Korean users.  Instead, the government will focus on keeping phones away from students. Extra staff will watch the hallways for Wi-Fi, LTE or 5G signals. What happens to people who cheat? The current war against cheating follows a confession that shocked the country. Han Won-seok, who scored highest on this year’s certified tax accountant exam, admitted on Sunday that he cheated.  Using a tiny camera, he photographed the questions, sent them to AI and then read the answers off his smartphone. He said he did the same thing in the first stage of the certified public accountant exam. People grew suspicious because his scores seemed too good to be true. In the April first stage, he got 159 of 160 questions right and scored 397.5 out of 400. He posted the results online before he confessed. Seoul police have opened an investigation, and the exam administrator said it will file a criminal complaint. Lee has ordered a reward of at least 100 million won (about $75,000) for each confirmed tip about cheating with advanced devices. The reward will be bigger if several people are involved. Officials are also thinking about lighter penalties for people who turn themselves in. Anyone caught cheating faces a ban of up to 10 years from the CSAT, plus criminal charges. Lee said using AI to get answers during the exam counts as obstructing official duties or business, which can bring up to five years in prison.  The government will also send its highest level of emergency phone alerts before and during the test. These alerts sound even on silenced phones, which could reveal students who kept their devices. Lee said that those thinking about cheating “should remember that every record is preserved.” His proposed law would apply the same rule to future national qualification exams, not just the CSAT. AI cheating has reportedly been a thing for at least two years, with 12 cases involving smart glasses this year alone. All 528 national qualification exam categories held this year are now under review. If you're reading this, you’re already ahead. Stay there with our newsletter.

Seoul to search AI chat logs after exam to catch cheaters

South Korea intends to look through AI chatbot records after its upcoming college entrance exam to find students who used AI to get answers.
The exam is on Thursday and the plan so far depends on an unfinished deal with four tech firms and on a law that has not even been written yet.
What is South Korea planning to do about AI cheating on its college exam?
Every year, South Korean students take a national college entrance test called the Suneung, also known as the College Scholastic Ability Test (CSAT). President Lee Jae Myung says the government will check AI chatbot records after this year’s test to find students who took photos of exam questions and asked AI for the answers.
He shared the plan on X and at an emergency video meeting he led from Manila, where he was attending a summit of Southeast Asian nations.
When a student sends a photo of a question to a chatbot, the image goes to the company’s servers and leaves a record of use. With this new rule, if investigators suspect a student of cheating, they can ask the company for that record.
Science Minister and Deputy Prime Minister Bae Kyung-hoon said four global tech companies have agreed to help when the South Korean government asks, but he admitted the companies have not agreed on how far that help will go.
The plan would require a new law capable of forcing AI companies to hand over records of illegal requests for exam answers. Lee has ordered that law, but it has not been passed.
The Korean government considered blocking the chatbots during the test but dropped the idea. China’s main AI services briefly refused to solve photos of exam questions during its college entrance exam in June 2025.
Deputy Prime Minister Bae said Korea cannot easily copy China’s approach, because it would require foreign AI firms to build a filter just for Korean users.
Instead, the government will focus on keeping phones away from students. Extra staff will watch the hallways for Wi-Fi, LTE or 5G signals.
What happens to people who cheat?
The current war against cheating follows a confession that shocked the country. Han Won-seok, who scored highest on this year’s certified tax accountant exam, admitted on Sunday that he cheated.
Using a tiny camera, he photographed the questions, sent them to AI and then read the answers off his smartphone. He said he did the same thing in the first stage of the certified public accountant exam.
People grew suspicious because his scores seemed too good to be true. In the April first stage, he got 159 of 160 questions right and scored 397.5 out of 400. He posted the results online before he confessed. Seoul police have opened an investigation, and the exam administrator said it will file a criminal complaint.
Lee has ordered a reward of at least 100 million won (about $75,000) for each confirmed tip about cheating with advanced devices. The reward will be bigger if several people are involved. Officials are also thinking about lighter penalties for people who turn themselves in.
Anyone caught cheating faces a ban of up to 10 years from the CSAT, plus criminal charges. Lee said using AI to get answers during the exam counts as obstructing official duties or business, which can bring up to five years in prison.
The government will also send its highest level of emergency phone alerts before and during the test. These alerts sound even on silenced phones, which could reveal students who kept their devices.
Lee said that those thinking about cheating “should remember that every record is preserved.” His proposed law would apply the same rule to future national qualification exams, not just the CSAT.
AI cheating has reportedly been a thing for at least two years, with 12 cases involving smart glasses this year alone. All 528 national qualification exam categories held this year are now under review.
If you're reading this, you’re already ahead. Stay there with our newsletter.
Best AI Video Generators in 2026: 11 Tools ComparedThere are many new versions of AI video tools being released, so it can be difficult to keep up with the latest. This article will give you an overview of 11 AI video tools and how they can be used. We’ve grouped the tools by function and provided information on what type of tool they are, main features, advantages, disadvantages and pricing. All tools below have genuine disadvantages listed and pricing details include plan, frequency and date checked. We’ve covered the drivers behind this market already, including Wan 3.0 providing 30-second video from spreadsheets, Gemini Omni paywall with free 1080p in Google Vids. This article will help you choose which tool best fits your needs. What changed recently: In early November, Google updated its AI video tools in Flow to use Gemini Omni 1.2 Flash. They can provide up to 15-second videos for Google AI subscribers. Sora was closed by OpenAI on 26th April 2026 and the API was closed on 24th September 2026. It is not included here. Kling 4.0 is now available to all users on the Kling site. Audio is also included. Pika has simplified their plan titles to Free, Starter, Creator and Fancy. The pricing page was last updated on 28th October 2026. Best AI Video Generators at a Glance Runway, Pika, Google Gemini Omni in Flow and Kling provide text-to-video and image-to-video. VEED and Pixlr provide editing tools with AI video generation. HeyGen and Synthesia provide presenter video, and Artlist provides licensed music and footage. Choose which best fits your needs. Tool Type Free tier Starting price (plan, billing, check date) Native audio Platforms Image-to-video Best for Runway Generator Yes: 125 one-time credits, watermark Standard, 12/mobilledannually(15 monthly), 625 credits; Nov 13, 2026 Varies by model Web, iOS, API Yes Directed shots and effects Pika Generator Yes: no watermark, no monthly credits (packs only), no commercial license Starter, 10/mo(8 billed annually), 600 credits; Nov 13, 2026 Varies by model Web, iOS Yes Social clips with many models in one place PixVerse Generator Yes: daily credits, watermark, up to 540p Standard, 10/mo(8 billed annually), 1,200 credits; Nov 13, 2026 Varies Web Yes Quick, low-cost social clips Leonardo.ai Generator and image platform Yes: 150 tokens a day Essential, 12/mo(10 billed yearly); Nov 13, 2026 Varies by model Web Yes Image-first creators adding video Google Veo and Gemini Omni Generator Limited Google AI Plus, $4.99/mo, 200 Flow credits; Nov 13, 2026 Yes (Omni) Gemini app; Flow on web Yes (reference images) Longer single shots with sound Kling Generator Basic plan; credits not listed Standard, 10/molistprice(6.99 first month), 660 credits; Nov 13, 2026 Yes (Kling 4.0) Web Yes Cinematic clips with synced audio VEED Editor Yes: 720p, watermark, 10-minute limit Lite, 12/mobilledannually(19 monthly); Nov 13, 2026 Varies Web Captions and social edits Pixlr Editor and generator Free editor; video credits Plus, 2.49/mo(1.99 billed yearly), 80 credits; Nov 13, 2026 Varies by model Web, desktop, iOS, Android Yes Photo editing plus quick clips HeyGen Avatar Yes: 3 videos a month, up to 1 minute Creator, 29/mo(24 billed annually), 600 credits; Nov 13, 2026 Yes (AI voice) Web Photo avatars only Presenter videos and translation Synthesia Avatar Yes: Basic plan, limited minutes Starter, $29/mo billed monthly, about 10 minutes; Nov 13, 2026 Yes (AI voice) Web No Training and internal video Artlist Stock, music and AI No permanent free plan AI plan, from $9.99/mo billed annually, 7,500 credits; Nov 13, 2026 Varies by model Web Yes Licensed music and footage What to Consider Before Choosing an AI Video Generator Below are 6 considerations for choosing an AI video generator. Text-to-video vs avatar vs editor vs stock/music Choosing a tool that fits your needs. These tools generate video from a prompt or image. Avatar tools will create a virtual presenter to read your script. Edits can be made to video with the addition of some now offering generation. Stock/Music offer licensed assets to be used with your AI video. Most disappointment comes from purchasing one type of tool when another is needed. Output quality, clip length and resolution AI video generation will always be short. Google Flow now provides up to 15-second videos for subscribers and 10 seconds for standard users. Many tools will only produce between 5-10 seconds of video. The resolution will depend on the model and will often be limited on free/entry level plans. Any longer than one video will require editing. Native audio Some models will produce video with audio, dialogue & effects. Some tools will produce silent video that will need music/voice added. As many of the tools (Runway, Pika, Leonardo.ai) have multiple models available, this will often vary by model. Credits, pricing and watermarks Most tools will use credits. Depending on the model, length & resolution, the cost per video will vary. Ensure you know if you will lose credits for failed or unwanted video. Will credits roll over? Are there watermarks on the free tier? You may be able to create a few premium videos for the same monthly cost. Commercial use rights and licensing Most tools will have a commercial licence dependent on your plan. Free plans may not allow commercial use even if there is no watermark. You may also need a licence for the video you upload, music or likeness. We recommend checking the terms page. Platforms and mobile apps Most tools will be a web app. If you want to use on mobile, ensure there is an iOS/Android app available on the official app store. Some tools will only provide some features and top up credits on the web. How This List Was Put Together The list was created by considering the 6 factors above. We checked the pricing and terms of each tool where possible. We also checked the release notes, help pages and public news. Plans were checked on November 13, 2026. We did not perform any hands-on testing of the tools, so pros and cons are based on the tool providers’ documentation and public reports. AI video tool pricing can change every few weeks. All prices should be considered a snapshot in time. Always check on the tool provider’s page. Annual pricing is shown as monthly. A few popular tools are not included (Luma, Invideo, Genmo, Adobe Firefly) and may be added in a refresh. Text-to-Video and Image-to-Video Generators These tools will generate new video based on your text prompt or image. The tools are only ordered from longest to shortest due to the complexity of the tools, not quality. Runway Runway Type: Generator (text-to-video, image-to-video, video-to-video) Key Features: Runway provides a wide range of models, including their own (Gen-4.5 and Aleph 2.0) alongside third-party models within one web app. It is well known for having creative controls and generating effects such as lighting, angle changes and object swapping. There is also a developer API available, charged via credits. Pros: Multiple model families within one app. Watermark removed with paid plans. 4K upscaling. Runway state has no non-commercial restrictions on content created. Cons: Only 125 one-off credits on the free plan. Watermark. Monthly credits do not roll over. Separate credit accounts for app and API. Paid third-party models use a lot of credits. Pricing: Free – $0. 125 one-off credits. Standard – $12/mobilledannually (15 monthly). 625 Credits Monthly. Pro – $28/mobilledannually (35 monthly). 2,250 Credits. Max – $76/mobilledannually (95 monthly). 9,500 Credits. Nov 13, 2026. Best For: Directed video and effects. Pika Pika Type: Generator (text-to-video, image-to-video, effects, audio) Key Features: Pika now provides a multi-model studio. It includes its own models (Pika H3 Flex and Pika 2.5) alongside third-party models (Seedance 2.5, Kling 4.0, Wan 3.0, Gemini Omni 1.1 Flash). Music, speech and sound effect studios are also available. Each plan includes all models. The Creator and Fancy plans also provide free daily Pika H3 Flex generations (limited-time offer). Pros: Wide range of models and audio tools. All on one credit account. No watermark on free plan exports. Commercial licence included with paid plans. Cons: No monthly credits on free plan. Only credit packs available. No commercial licence. Many credits are required for paid models. 122 Credits for a 5-second Seedance 2.5 at 720p. Starter will provide ~4 videos. Top-up credits are only available on the web app and MCP. Not API and iOS. Priority Queue is only available on the Fancy plan. Pricing: Free – $0. Only credit packs are available. Starter – $10/mo (8 annually). 600 Credits. Creator – $35/mo (28 annually). 2,600 Credits. Fancy – From $95/mo (76 annually). 8,550+ Credits. Updated 28th October 2026. Best For: Social video and effects. With lots of models. PixVerse PixVerse Type: Generator (text-to-video, image-to-video) Key Features: PixVerse is a web-based video tool with a free plan and daily credits. The blog on their site appears first on Google for this search and PixVerse is listed first. All points below are based on statements made by the vendor and reported by third parties. Paid plans remove the watermark and increase resolution. Higher plans provide 4K. Pros: Free plan with daily credits. Text-to-video and image-to-video. Low-cost entry plan. No watermark on paid plans. Cons: Watermarked and limited to 540p on the free plan. Cannot be used commercially. Daily credits on free plan. Not bankable. Standard plan limited to 720p. Number of free credits varies between reports (30 and 60). Check prior to using. Can drift with long/complex prompts. Pricing: Basic – Free. Standard – 10/mo (8 annually). 1,200 monthly credits.Pro–~30/mo. Nov 13, 2026. Best For: Fast and cheap social video. Leonardo.ai Leonardo.ai Type: Generator in an image and design software Key Features: Leonardo is an image-focused tool owned by Canva. Video can be created using its own Motion models and third-party models. All are paid using tokens. On the pricing page, the Ultimate plan provides unlimited relaxed video for select models (Motion 2.0 and Motion 2.0 Fast). Other models use tokens. Pros: All-in-one Image, video, and Design tool. Free plan provides 150 tokens daily. Ultimate plan provides relaxed video. Great for regular volume. Cons: Video is not primary. Not clearly published how many tokens are required per video. Content created is public on the free plan. Unlimited only for select models. Pricing: Free – $0. 150 daily tokens. Essentials – $12/mo (10 annually). Premium – $30/mo. Ultimate – $60/mo (48 annually). Nov 13, 2026. Monthly tokens vary. Best For: Image-focused creators wanting to add video. Google Veo and Gemini Omni Google Veo and Gemini Omni Type: Generator (text, image, and video inputs) Key Features: Google has moved its consumer video to Gemini Omni. The Gemini app provides video generation using Omni. Within Flow, Google’s filmmaking tool, Omni and Veo 3.1 are available. Omni provides video with audio. We have covered the product side of Gemini Omni, with the paywall and free 1080p in Google Vids. Pros: Audio provided natively. Up to 15-second video in Flow for subscribers. Available as part of Google AI plans and storage. Cons: All plans have limited credits. Veo Extend has been retired. Some old models in Flow will be retired in December. The latest model in Flow is only available via the web. Pricing: Google AI Plus – $4.99/mo. 200 Flow Credits. Google AI Pro – $19.99/mo. 1,000 Credits. Google AI Ultra – 10,000 and 25,000 Credits. Nov 13, 2026. Best For: Longer single video with audio. Kling Kling Type: Generator (text-to-video, image-to-video) Key Features: Kling, by Kuaishou, has made Kling 4.0 available to all users. Audio is also available. Kling models can also be found within Runway, Pika, Pixlr & Leonardo.ai. Pros: Audio is available in 4.0. Free Basic Plan. Available on multiple other platforms. Cons: Kling has not listed how many credits are required for one Kling 4.0 video. No free credits listed on the Basic Plan. Many fake Kling 4 sites. Use the official site. Pricing: Standard – $10/molist (6.99 first month, $8.80 after). 660 Credits. Nov 13, 2026. Best For: Cinematic video with audio. AI Video Editors and All-in-One Suites Which is the best AI Video Editor for you will depend on whether you wish to edit video or use a prompt. These tools provide generation within their editing tools. CapCut also provides AI video generation. See FAQ for free plan limitations. VEED Type: Browser-based editor with AI capabilities Key Features: VEED provides video editing, auto captioning, translation and social exporting. Text-to-Video and AI Avatar tools are reported to be only available on the Pro plan. The free plan does not provide AI generation. Pros: No installation required. Fast captioning. Lite plan removes watermark and provides 1080p. Pro plan provides 4K and generation tools. Cons: All exports are watermarked. Limited to 10-minute videos. 720p. Generation tools on higher plan. Not a proper professional timeline editor, according to reviewers. Price: Free – $0. Lite – $12/mobilledannually (19 monthly). Pro – $24/mobilledannually (49 monthly). Nov 13, 2026. Some pages show different prices. Best For: Captioning, tidying and social. Pixlr Type: Photo Editor and Image Generator (With AI Video Generation) Key Features: Pixlr is best known as an image editor, but on their site they have an AI Video Generator. This can generate video from text/images using models such as Kling, Veo and Seedance. They also have an audio generator. Their subscriptions can be used on the web, desktop and mobile. Pros: All-in-one editor, image generator and video. All in browser. Low-cost entry plan. No expiry on credit packs. Cons: Video is just one module. Credits reset monthly on subscriptions. If cancelled, they will go to zero. They state AI services are not guaranteed. No video available on the free plan. Pricing: Plus – $2.49/mo (1.99 annually). 80 Credits (~4 fast videos). Premium – $9.99/mo (7.99 annually). 1,000 Credits (~50 fast videos). All video models. Ultra – From $24.99/mo (19.99 annually). Nov 13, 2026. Best For: Photo editing and quick video. Avatar and Presenter Video Avatar tools will take your script and create a presenter. They are ideal for explainer, training and sales videos but not cinematic. HeyGen Type: Avatar and translation video Key Features: HeyGen provides avatar tools with lip sync, gestures and video translation. On their pricing page, Avatar IV and V videos are $20 per minute. HeyGen also appears first on their own product for this search, so consider this when reading. Pros: High-quality avatars, translation and a free plan to try. Creator plan removes watermark and provides up to 30 minutes of video. Cons: Credits run out quickly. 600 credits will provide ~30 minutes of Avatar IV/V video. Costs credits to re-render. Extra video avatars cost extra. Pricing: Free – $0. 3 videos monthly (up to 1 minute). Creator – $29/mo (24 annually). 600 Credits. Pro – From $49/mo. 1,000 Credits. Nov 13, 2026. Best For: Presenter video and translation. Synthesia Type: Avatar and training video Key Features: Synthesia is focused on training and corporate video. They provide collaborative features and exports for LMS (SCORM). Pros: Ideal for training and internal communications. Free plan to try. Cons: All self-serve plans have limited minutes per month. Lower plans have limited avatars. Annual discounts and minutes vary between sources. Cost: Basic – $0. Starter – $29/mo monthly. ~10 minutes per month. Creator – $89/mo monthly. ~30 minutes. Nov 13, 2026. Best For: Training and internal video. Stock, Music and Assets for AI Video AI video is rarely provided on its own. Music, sound effects and footage can be used to fill the gaps. Artlist Type: Stock, music, and assets platform with AI Suite Key Features: Artlist is primarily a music, footage and templates library. They also have an AI suite on their site. This provides video, image, music and voice generation. They also provide image-to-video, video-to-video, access to over 100 models and Artlist Studio. The Max plan includes stock and AI. Pros: Music, footage and AI generation all in one license. Footage up to 8K. Premiere Pro plugins. AI plans provide commercial use on any platform. Cons: Not primarily a generator. AI credits are separate from stock. Stock licensing is dependent on channel and use. No permanently free plan. Plan names and pricing have changed. Pricing: AI Plans – From $9.99/mobilledannually (15.99 monthly). 7,500 Credits. Up to 23.99/moannually. 40,000 credits.Footage & Templates–~31.99/mo annually. Nov 13, 2026. Best For: Licensed music and footage with AI. Best AI Video Generator by Use Case These recommendations only relate to tools listed above. Based on documented features. Best AI video generator for YouTube Longer videos will help. Gemini Omni in Flow provides up to 15 seconds of video with audio. VEED provides captioning and editing. Artlist provides unlimited single videos on a Pro licence. HeyGen is great for talking head explainers. Best AI video generator for TikTok and short-form Pika provides effects and multiple models. PixVerse provides cheap video. VEED provides captioning and social. Pixlr provides mobile apps to quickly edit. Best AI music video generator While there is no one-stop-shop tool for creating full music videos, Pika has music and sound studios; Artlist provides AI music plus a music library; Pixlr has an audio generator list. Audio-enabled models such as Gemini Omni and Kling 4.0 will have audio within the video clips. However, these clips are always short, so you will need to edit them together using a video editor such as VEED. Best AI image-to-video generator Runway, Pika, Kling, Pixlr, Leonardo.ai and Artlist all have image-to-video capabilities. Gemini Omni has reference images. However, the cost per clip for stills-to-video on the premium models is not cheap. Best AI animation and anime generator None of the vendors offer verified specific anime modes. Pika has effect tools and Leonardo.ai has a model/style library. Best AI generator for short films Runway and Google Flow provide the closest solution, as they have director controls and film capabilities with 15-second shot extension. You will need to stitch and have a separate music/sound pass. Common Pitfalls to Avoid Wasting credits by re-rolling. Most videos will require multiple attempts. Divide your video credits up by how much your model is charging per clip. Free versions. Free versions can vary between vendors. Some will include watermarks or resolution limitations. Pika’s free version doesn’t have a watermark but no monthly credits. Some plans will restrict you from using the free version commercially. Pika’s free version doesn’t have a commercial licence, and multiple tools don’t allow you to use the free output commercially. Short clip lengths. These only allow for 5- 15-second clips, so you will need to stitch longer videos. Versions can change rapidly. This can happen within days. Check the vendor’s webpage and again at checkout. Fake sites. Some, like Kling 4, have fake sites that charge you for access. Cryptopolitan also talks about labelling AI-generated videos, such as Google’s SynthID detection tool. Final Verdict: How to Choose the Best AI Video Generator It is impossible to say which is the best AI video generator, as they are designed to handle different tasks. When creating cinematic text-to-video, compare Runway, Pika, Gemini Omni (Flow), and Kling on cost per video clip (credits), commercial rights and audio. If you require tools for presentations, then HeyGen and Synthesia would be the best choices. If you want to edit videos and add subtitles, then use VEED or Pixlr. If you need licensed music/footage, then Artlist will be your choice.

Best AI Video Generators in 2026: 11 Tools Compared

There are many new versions of AI video tools being released, so it can be difficult to keep up with the latest. This article will give you an overview of 11 AI video tools and how they can be used. We’ve grouped the tools by function and provided information on what type of tool they are, main features, advantages, disadvantages and pricing.
All tools below have genuine disadvantages listed and pricing details include plan, frequency and date checked. We’ve covered the drivers behind this market already, including Wan 3.0 providing 30-second video from spreadsheets, Gemini Omni paywall with free 1080p in Google Vids. This article will help you choose which tool best fits your needs.
What changed recently:
In early November, Google updated its AI video tools in Flow to use Gemini Omni 1.2 Flash. They can provide up to 15-second videos for Google AI subscribers.
Sora was closed by OpenAI on 26th April 2026 and the API was closed on 24th September 2026. It is not included here.
Kling 4.0 is now available to all users on the Kling site. Audio is also included.
Pika has simplified their plan titles to Free, Starter, Creator and Fancy. The pricing page was last updated on 28th October 2026.
Best AI Video Generators at a Glance
Runway, Pika, Google Gemini Omni in Flow and Kling provide text-to-video and image-to-video. VEED and Pixlr provide editing tools with AI video generation. HeyGen and Synthesia provide presenter video, and Artlist provides licensed music and footage. Choose which best fits your needs.
Tool Type Free tier Starting price (plan, billing, check date) Native audio Platforms Image-to-video Best for Runway Generator Yes: 125 one-time credits, watermark Standard, 12/mobilledannually(15 monthly), 625 credits; Nov 13, 2026 Varies by model Web, iOS, API Yes Directed shots and effects Pika Generator Yes: no watermark, no monthly credits (packs only), no commercial license Starter, 10/mo(8 billed annually), 600 credits; Nov 13, 2026 Varies by model Web, iOS Yes Social clips with many models in one place PixVerse Generator Yes: daily credits, watermark, up to 540p Standard, 10/mo(8 billed annually), 1,200 credits; Nov 13, 2026 Varies Web Yes Quick, low-cost social clips Leonardo.ai Generator and image platform Yes: 150 tokens a day Essential, 12/mo(10 billed yearly); Nov 13, 2026 Varies by model Web Yes Image-first creators adding video Google Veo and Gemini Omni Generator Limited Google AI Plus, $4.99/mo, 200 Flow credits; Nov 13, 2026 Yes (Omni) Gemini app; Flow on web Yes (reference images) Longer single shots with sound Kling Generator Basic plan; credits not listed Standard, 10/molistprice(6.99 first month), 660 credits; Nov 13, 2026 Yes (Kling 4.0) Web Yes Cinematic clips with synced audio VEED Editor Yes: 720p, watermark, 10-minute limit Lite, 12/mobilledannually(19 monthly); Nov 13, 2026 Varies Web Captions and social edits Pixlr Editor and generator Free editor; video credits Plus, 2.49/mo(1.99 billed yearly), 80 credits; Nov 13, 2026 Varies by model Web, desktop, iOS, Android Yes Photo editing plus quick clips HeyGen Avatar Yes: 3 videos a month, up to 1 minute Creator, 29/mo(24 billed annually), 600 credits; Nov 13, 2026 Yes (AI voice) Web Photo avatars only Presenter videos and translation Synthesia Avatar Yes: Basic plan, limited minutes Starter, $29/mo billed monthly, about 10 minutes; Nov 13, 2026 Yes (AI voice) Web No Training and internal video Artlist Stock, music and AI No permanent free plan AI plan, from $9.99/mo billed annually, 7,500 credits; Nov 13, 2026 Varies by model Web Yes Licensed music and footage
What to Consider Before Choosing an AI Video Generator
Below are 6 considerations for choosing an AI video generator.
Text-to-video vs avatar vs editor vs stock/music
Choosing a tool that fits your needs. These tools generate video from a prompt or image. Avatar tools will create a virtual presenter to read your script. Edits can be made to video with the addition of some now offering generation. Stock/Music offer licensed assets to be used with your AI video. Most disappointment comes from purchasing one type of tool when another is needed.
Output quality, clip length and resolution
AI video generation will always be short. Google Flow now provides up to 15-second videos for subscribers and 10 seconds for standard users. Many tools will only produce between 5-10 seconds of video. The resolution will depend on the model and will often be limited on free/entry level plans. Any longer than one video will require editing.
Native audio
Some models will produce video with audio, dialogue & effects. Some tools will produce silent video that will need music/voice added. As many of the tools (Runway, Pika, Leonardo.ai) have multiple models available, this will often vary by model.
Credits, pricing and watermarks
Most tools will use credits. Depending on the model, length & resolution, the cost per video will vary. Ensure you know if you will lose credits for failed or unwanted video. Will credits roll over? Are there watermarks on the free tier? You may be able to create a few premium videos for the same monthly cost.
Commercial use rights and licensing
Most tools will have a commercial licence dependent on your plan. Free plans may not allow commercial use even if there is no watermark. You may also need a licence for the video you upload, music or likeness. We recommend checking the terms page.
Platforms and mobile apps
Most tools will be a web app. If you want to use on mobile, ensure there is an iOS/Android app available on the official app store. Some tools will only provide some features and top up credits on the web.
How This List Was Put Together
The list was created by considering the 6 factors above. We checked the pricing and terms of each tool where possible. We also checked the release notes, help pages and public news. Plans were checked on November 13, 2026. We did not perform any hands-on testing of the tools, so pros and cons are based on the tool providers’ documentation and public reports.
AI video tool pricing can change every few weeks. All prices should be considered a snapshot in time. Always check on the tool provider’s page. Annual pricing is shown as monthly. A few popular tools are not included (Luma, Invideo, Genmo, Adobe Firefly) and may be added in a refresh.
Text-to-Video and Image-to-Video Generators
These tools will generate new video based on your text prompt or image. The tools are only ordered from longest to shortest due to the complexity of the tools, not quality.
Runway
Runway
Type: Generator (text-to-video, image-to-video, video-to-video)
Key Features: Runway provides a wide range of models, including their own (Gen-4.5 and Aleph 2.0) alongside third-party models within one web app. It is well known for having creative controls and generating effects such as lighting, angle changes and object swapping. There is also a developer API available, charged via credits.
Pros:
Multiple model families within one app.
Watermark removed with paid plans. 4K upscaling.
Runway state has no non-commercial restrictions on content created.
Cons:
Only 125 one-off credits on the free plan. Watermark.
Monthly credits do not roll over.
Separate credit accounts for app and API.
Paid third-party models use a lot of credits.
Pricing: Free – $0. 125 one-off credits. Standard – $12/mobilledannually (15 monthly). 625 Credits Monthly. Pro – $28/mobilledannually (35 monthly). 2,250 Credits. Max – $76/mobilledannually (95 monthly). 9,500 Credits. Nov 13, 2026.
Best For: Directed video and effects.
Pika
Pika
Type: Generator (text-to-video, image-to-video, effects, audio)
Key Features: Pika now provides a multi-model studio. It includes its own models (Pika H3 Flex and Pika 2.5) alongside third-party models (Seedance 2.5, Kling 4.0, Wan 3.0, Gemini Omni 1.1 Flash). Music, speech and sound effect studios are also available. Each plan includes all models. The Creator and Fancy plans also provide free daily Pika H3 Flex generations (limited-time offer).
Pros:
Wide range of models and audio tools. All on one credit account.
No watermark on free plan exports.
Commercial licence included with paid plans.
Cons:
No monthly credits on free plan. Only credit packs available. No commercial licence.
Many credits are required for paid models. 122 Credits for a 5-second Seedance 2.5 at 720p. Starter will provide ~4 videos.
Top-up credits are only available on the web app and MCP. Not API and iOS.
Priority Queue is only available on the Fancy plan.
Pricing: Free – $0. Only credit packs are available. Starter – $10/mo (8 annually). 600 Credits. Creator – $35/mo (28 annually). 2,600 Credits. Fancy – From $95/mo (76 annually). 8,550+ Credits. Updated 28th October 2026.
Best For: Social video and effects. With lots of models.
PixVerse
PixVerse
Type: Generator (text-to-video, image-to-video)
Key Features: PixVerse is a web-based video tool with a free plan and daily credits. The blog on their site appears first on Google for this search and PixVerse is listed first. All points below are based on statements made by the vendor and reported by third parties. Paid plans remove the watermark and increase resolution. Higher plans provide 4K.
Pros:
Free plan with daily credits. Text-to-video and image-to-video.
Low-cost entry plan.
No watermark on paid plans.
Cons:
Watermarked and limited to 540p on the free plan. Cannot be used commercially.
Daily credits on free plan. Not bankable.
Standard plan limited to 720p.
Number of free credits varies between reports (30 and 60). Check prior to using.
Can drift with long/complex prompts.
Pricing: Basic – Free. Standard – 10/mo (8 annually). 1,200 monthly credits.Pro–~30/mo. Nov 13, 2026.
Best For: Fast and cheap social video.
Leonardo.ai
Leonardo.ai
Type: Generator in an image and design software
Key Features: Leonardo is an image-focused tool owned by Canva. Video can be created using its own Motion models and third-party models. All are paid using tokens. On the pricing page, the Ultimate plan provides unlimited relaxed video for select models (Motion 2.0 and Motion 2.0 Fast). Other models use tokens.
Pros:
All-in-one Image, video, and Design tool.
Free plan provides 150 tokens daily.
Ultimate plan provides relaxed video. Great for regular volume.
Cons:
Video is not primary.
Not clearly published how many tokens are required per video.
Content created is public on the free plan.
Unlimited only for select models.
Pricing: Free – $0. 150 daily tokens. Essentials – $12/mo (10 annually). Premium – $30/mo. Ultimate – $60/mo (48 annually). Nov 13, 2026. Monthly tokens vary.
Best For: Image-focused creators wanting to add video.
Google Veo and Gemini Omni
Google Veo and Gemini Omni
Type: Generator (text, image, and video inputs)
Key Features: Google has moved its consumer video to Gemini Omni. The Gemini app provides video generation using Omni. Within Flow, Google’s filmmaking tool, Omni and Veo 3.1 are available. Omni provides video with audio. We have covered the product side of Gemini Omni, with the paywall and free 1080p in Google Vids.
Pros:
Audio provided natively.
Up to 15-second video in Flow for subscribers.
Available as part of Google AI plans and storage.
Cons:
All plans have limited credits.
Veo Extend has been retired. Some old models in Flow will be retired in December.
The latest model in Flow is only available via the web.
Pricing: Google AI Plus – $4.99/mo. 200 Flow Credits. Google AI Pro – $19.99/mo. 1,000 Credits. Google AI Ultra – 10,000 and 25,000 Credits. Nov 13, 2026.
Best For: Longer single video with audio.
Kling
Kling
Type: Generator (text-to-video, image-to-video)
Key Features: Kling, by Kuaishou, has made Kling 4.0 available to all users. Audio is also available. Kling models can also be found within Runway, Pika, Pixlr & Leonardo.ai.
Pros:
Audio is available in 4.0.
Free Basic Plan.
Available on multiple other platforms.
Cons:
Kling has not listed how many credits are required for one Kling 4.0 video.
No free credits listed on the Basic Plan.
Many fake Kling 4 sites. Use the official site.
Pricing: Standard – $10/molist (6.99 first month, $8.80 after). 660 Credits. Nov 13, 2026.
Best For: Cinematic video with audio.
AI Video Editors and All-in-One Suites
Which is the best AI Video Editor for you will depend on whether you wish to edit video or use a prompt. These tools provide generation within their editing tools. CapCut also provides AI video generation. See FAQ for free plan limitations.
VEED
Type: Browser-based editor with AI capabilities
Key Features: VEED provides video editing, auto captioning, translation and social exporting. Text-to-Video and AI Avatar tools are reported to be only available on the Pro plan. The free plan does not provide AI generation.
Pros:
No installation required. Fast captioning.
Lite plan removes watermark and provides 1080p.
Pro plan provides 4K and generation tools.
Cons:
All exports are watermarked. Limited to 10-minute videos. 720p.
Generation tools on higher plan.
Not a proper professional timeline editor, according to reviewers.
Price: Free – $0. Lite – $12/mobilledannually (19 monthly). Pro – $24/mobilledannually (49 monthly). Nov 13, 2026. Some pages show different prices.
Best For: Captioning, tidying and social.
Pixlr
Type: Photo Editor and Image Generator (With AI Video Generation)
Key Features: Pixlr is best known as an image editor, but on their site they have an AI Video Generator. This can generate video from text/images using models such as Kling, Veo and Seedance. They also have an audio generator. Their subscriptions can be used on the web, desktop and mobile.
Pros:
All-in-one editor, image generator and video. All in browser.
Low-cost entry plan.
No expiry on credit packs.
Cons:
Video is just one module.
Credits reset monthly on subscriptions. If cancelled, they will go to zero.
They state AI services are not guaranteed.
No video available on the free plan.
Pricing: Plus – $2.49/mo (1.99 annually). 80 Credits (~4 fast videos). Premium – $9.99/mo (7.99 annually). 1,000 Credits (~50 fast videos). All video models. Ultra – From $24.99/mo (19.99 annually). Nov 13, 2026.
Best For: Photo editing and quick video.
Avatar and Presenter Video
Avatar tools will take your script and create a presenter. They are ideal for explainer, training and sales videos but not cinematic.
HeyGen
Type: Avatar and translation video
Key Features: HeyGen provides avatar tools with lip sync, gestures and video translation. On their pricing page, Avatar IV and V videos are $20 per minute. HeyGen also appears first on their own product for this search, so consider this when reading.
Pros:
High-quality avatars, translation and a free plan to try.
Creator plan removes watermark and provides up to 30 minutes of video.
Cons:
Credits run out quickly. 600 credits will provide ~30 minutes of Avatar IV/V video.
Costs credits to re-render.
Extra video avatars cost extra.
Pricing: Free – $0. 3 videos monthly (up to 1 minute). Creator – $29/mo (24 annually). 600 Credits. Pro – From $49/mo. 1,000 Credits. Nov 13, 2026.
Best For: Presenter video and translation.
Synthesia
Type: Avatar and training video
Key Features: Synthesia is focused on training and corporate video. They provide collaborative features and exports for LMS (SCORM).
Pros:
Ideal for training and internal communications.
Free plan to try.
Cons:
All self-serve plans have limited minutes per month.
Lower plans have limited avatars.
Annual discounts and minutes vary between sources.
Cost: Basic – $0. Starter – $29/mo monthly. ~10 minutes per month. Creator – $89/mo monthly. ~30 minutes. Nov 13, 2026.
Best For: Training and internal video.
Stock, Music and Assets for AI Video
AI video is rarely provided on its own. Music, sound effects and footage can be used to fill the gaps.
Artlist
Type: Stock, music, and assets platform with AI Suite
Key Features: Artlist is primarily a music, footage and templates library. They also have an AI suite on their site. This provides video, image, music and voice generation. They also provide image-to-video, video-to-video, access to over 100 models and Artlist Studio. The Max plan includes stock and AI.
Pros:
Music, footage and AI generation all in one license.
Footage up to 8K. Premiere Pro plugins.
AI plans provide commercial use on any platform.
Cons:
Not primarily a generator. AI credits are separate from stock.
Stock licensing is dependent on channel and use.
No permanently free plan.
Plan names and pricing have changed.
Pricing: AI Plans – From $9.99/mobilledannually (15.99 monthly). 7,500 Credits. Up to 23.99/moannually. 40,000 credits.Footage & Templates–~31.99/mo annually. Nov 13, 2026.
Best For: Licensed music and footage with AI.
Best AI Video Generator by Use Case
These recommendations only relate to tools listed above. Based on documented features.
Best AI video generator for YouTube
Longer videos will help. Gemini Omni in Flow provides up to 15 seconds of video with audio. VEED provides captioning and editing. Artlist provides unlimited single videos on a Pro licence. HeyGen is great for talking head explainers.
Best AI video generator for TikTok and short-form
Pika provides effects and multiple models. PixVerse provides cheap video. VEED provides captioning and social. Pixlr provides mobile apps to quickly edit.
Best AI music video generator
While there is no one-stop-shop tool for creating full music videos, Pika has music and sound studios; Artlist provides AI music plus a music library; Pixlr has an audio generator list. Audio-enabled models such as Gemini Omni and Kling 4.0 will have audio within the video clips. However, these clips are always short, so you will need to edit them together using a video editor such as VEED.
Best AI image-to-video generator
Runway, Pika, Kling, Pixlr, Leonardo.ai and Artlist all have image-to-video capabilities. Gemini Omni has reference images. However, the cost per clip for stills-to-video on the premium models is not cheap.
Best AI animation and anime generator
None of the vendors offer verified specific anime modes. Pika has effect tools and Leonardo.ai has a model/style library.
Best AI generator for short films
Runway and Google Flow provide the closest solution, as they have director controls and film capabilities with 15-second shot extension. You will need to stitch and have a separate music/sound pass.
Common Pitfalls to Avoid
Wasting credits by re-rolling. Most videos will require multiple attempts. Divide your video credits up by how much your model is charging per clip.
Free versions. Free versions can vary between vendors. Some will include watermarks or resolution limitations. Pika’s free version doesn’t have a watermark but no monthly credits.
Some plans will restrict you from using the free version commercially. Pika’s free version doesn’t have a commercial licence, and multiple tools don’t allow you to use the free output commercially.
Short clip lengths. These only allow for 5- 15-second clips, so you will need to stitch longer videos.
Versions can change rapidly. This can happen within days. Check the vendor’s webpage and again at checkout.
Fake sites. Some, like Kling 4, have fake sites that charge you for access.
Cryptopolitan also talks about labelling AI-generated videos, such as Google’s SynthID detection tool.
Final Verdict: How to Choose the Best AI Video Generator
It is impossible to say which is the best AI video generator, as they are designed to handle different tasks. When creating cinematic text-to-video, compare Runway, Pika, Gemini Omni (Flow), and Kling on cost per video clip (credits), commercial rights and audio. If you require tools for presentations, then HeyGen and Synthesia would be the best choices. If you want to edit videos and add subtitles, then use VEED or Pixlr. If you need licensed music/footage, then Artlist will be your choice.
Japan’s first yen stablecoin issuer suspends app indefinitely after string of service haltsJPYC announced on Wednesday that its apps would be taken offline indefinitely for unscheduled maintenance. The issuer of Japan’s first regulated yen stablecoin did not commit to a restart date and did not mention any reason for the downtime.  News of the service outage arrives barely one week after CEO Noritaka Okabe named the yen stablecoin his company issues as the users’ alternative when banks face their own issues.  JPYC runs the first yen stablecoin approved to operate inside Japan. Service outages affecting its apps reach the users and businesses that need to mint, redeem or move the token. Is JPYC down?  JPYC Co. confirmed in a notice on Wednesday that apps remain down due to unscheduled maintenance. According to the English translation of the message written in Japanese and posted on its official X account, the yen stablecoin issuer apologized to users “for the prolonged service outage.”  Translation of JPYC’s notice about app downtime. Source: @jpyc_official via X/Twitter However, JPYC did not specify a date or time when the affected apps could come back online. The notice did not name a cause for the downtime either. The JPYC service outage on November 18 is the second to hit the stablecoin issuer this month, following reports of maintenance on its issuance and redemption platform, JPYC EX, roughly one week ago.  Those JPYC EX maintenances, which were also described as unscheduled, came after partner bank, GMO Aozora Net Bank, suspended its service twice between November 10 and 11 after detecting two rounds of unauthorized external communication against its bank API. JPYC EX is the gateway between the JPYC stablecoin and the Japanese yen.  Binance Japan, BitTrade and the accounting software firm freee also had to take service offline as GMO Aozora Net Bank managed the security incident. CEO Okabe pits company against banks  JPYC taking its apps offline lands barely one week after Okabe pitched his company’s stablecoin as the economic lifeline for Japanese bank customers who would otherwise be stranded during bank outages.  According to Cryptopolitan’s coverage of the CEO’s November 11 X post, Okabe argued that the more bank rails go dark, the more valuable his company’s stablecoin becomes, because traditional payment systems do not affect public blockchains.  Okabe also noted the boon to the yen’s overall resilience as the economy can continue to run on JPYC during outages and then reconnect when the banks resolve their issues.  In late October, Okabe aimed at legacy finance, writing that “Existing finance doesn’t have long” following reports of unauthorized access at financial institutions, adding that “The lifespan of fiat currency is also uncertain.” JPYC has had a stuttering start since its October 2025 launch as the world’s first fully regulated yen-pegged stablecoin. It operates across Ethereum, Polygon, Avalanche and Kaia. Okabe conceded on October 27, 2026, the stablecoin’s first anniversary, that it had not met its initial circulation targets, per Cryptopolitan. However, he’s more optimistic of the future, pinning year-two growth on partnerships, including a logistics deal with AZ-COM Maruwa to pay 2,300 carriers and an October 20 memorandum with SBJ Bank and LINE NEXT. Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free.

Japan’s first yen stablecoin issuer suspends app indefinitely after string of service halts

JPYC announced on Wednesday that its apps would be taken offline indefinitely for unscheduled maintenance. The issuer of Japan’s first regulated yen stablecoin did not commit to a restart date and did not mention any reason for the downtime.
News of the service outage arrives barely one week after CEO Noritaka Okabe named the yen stablecoin his company issues as the users’ alternative when banks face their own issues.
JPYC runs the first yen stablecoin approved to operate inside Japan. Service outages affecting its apps reach the users and businesses that need to mint, redeem or move the token.
Is JPYC down?
JPYC Co. confirmed in a notice on Wednesday that apps remain down due to unscheduled maintenance. According to the English translation of the message written in Japanese and posted on its official X account, the yen stablecoin issuer apologized to users “for the prolonged service outage.”
Translation of JPYC’s notice about app downtime. Source: @jpyc_official via X/Twitter
However, JPYC did not specify a date or time when the affected apps could come back online. The notice did not name a cause for the downtime either.
The JPYC service outage on November 18 is the second to hit the stablecoin issuer this month, following reports of maintenance on its issuance and redemption platform, JPYC EX, roughly one week ago.
Those JPYC EX maintenances, which were also described as unscheduled, came after partner bank, GMO Aozora Net Bank, suspended its service twice between November 10 and 11 after detecting two rounds of unauthorized external communication against its bank API.
JPYC EX is the gateway between the JPYC stablecoin and the Japanese yen.
Binance Japan, BitTrade and the accounting software firm freee also had to take service offline as GMO Aozora Net Bank managed the security incident.
CEO Okabe pits company against banks
JPYC taking its apps offline lands barely one week after Okabe pitched his company’s stablecoin as the economic lifeline for Japanese bank customers who would otherwise be stranded during bank outages.
According to Cryptopolitan’s coverage of the CEO’s November 11 X post, Okabe argued that the more bank rails go dark, the more valuable his company’s stablecoin becomes, because traditional payment systems do not affect public blockchains.
Okabe also noted the boon to the yen’s overall resilience as the economy can continue to run on JPYC during outages and then reconnect when the banks resolve their issues.
In late October, Okabe aimed at legacy finance, writing that “Existing finance doesn’t have long” following reports of unauthorized access at financial institutions, adding that “The lifespan of fiat currency is also uncertain.”
JPYC has had a stuttering start since its October 2025 launch as the world’s first fully regulated yen-pegged stablecoin. It operates across Ethereum, Polygon, Avalanche and Kaia.
Okabe conceded on October 27, 2026, the stablecoin’s first anniversary, that it had not met its initial circulation targets, per Cryptopolitan. However, he’s more optimistic of the future, pinning year-two growth on partnerships, including a logistics deal with AZ-COM Maruwa to pay 2,300 carriers and an October 20 memorandum with SBJ Bank and LINE NEXT.
Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free.
Masayoshi Son calls Anthropic miss a 'mistake' a day before its IPOSoftBank’s CEO Masayoshi Son has rued missing the chance to get in on Anthropic at an earlier stage on Wednesday, one day before the Claude maker debuts on the Nasdaq after its November 19 IPO.  Son, whose SoftBank is heavily invested in OpenAI, admitted that not having stakes in the AI lab hitting the market tomorrow is one of his investment “mistakes.” OpenAI, on the other hand, has pushed its own IPO into 2027, delaying Son and SoftBank’s cashout on their investment.  Did SoftBank invest in Anthropic? Speaking at Nikkei’s “Dialogues on the Future” conference in Tokyo, SoftBank CEO Masayoshi Son talked about his firm’s all-in investment strategy in Sam Altman’s OpenAI, while skipping out on Dario Amodei’s Anthropic entirely.  Notably, Son admitted the difficulty of deciding which AI developer to hitch their wagon to in the AI race. His comments came as the OpenAI wagon that SoftBank jumped on continues to delay its IPO, while Anthropic, the company they have no stake in, is due to go public on Thursday under the ticker ANT on Nasdaq. Despite the apparent lag, the SoftBank founder has not cooled on OpenAI, telling the forum that he measures success by how his firm’s investments create the future rather than by making money. As of this report, SoftBank has committed up to $64.6 billion to OpenAI for a cumulative stake of roughly 13% held through Vision Fund 2. The final $10 billion tranche of a $30 billion follow-on investment in OpenAI was completed on October 1.  Who missed out on investing in Anthropic? Son joins a list of publicly tracked big-name investors who will not hold any stakes in Anthropic when it debuts on the Nasdaq, either by completely passing up early investment rounds, walking away or losing their stakes before the IPO.  Sam Bankman-Fried (SBF) lands in the latter category. The disgraced FTX founder, who is currently serving a 25-year sentence at the FCI Terminal Island low-security federal prison in San Pedro, Los Angeles, had a 13.5% stake in Anthropic via a late 2021 $500 million investment.  However, the bankruptcy estate sold that Anthropic position for $1.3 billion in 2024 in a flash sale to repay FTX customers after the exchange collapsed in November 2022 due to an $8 billion hole in customer accounts.  Former FTX executives Caroline Ellison and Nishad Singh also forfeited stakes to the US government. Those stakes netted $627 million when they were sold. The proceeds from that sale were directed back to the FTX estate. The other parties in the Anthropic stakes sale transactions are due for big paydays. One of them, Jane Street, will see the $1.34 billion it spent on FTX-linked Anthropic shares explode to be worth more than $35 billion upon stock listing. OpenAI’s round shrinks as Anthropic races to list The IPO timing lag is not the only point of contrast in the OpenAI versus Anthropic comparison. Anthropic priced its offering in a range of $70 to $78 a share, valuing the Claude maker at up to roughly $1.4 trillion, Fortune reported. The company booked $17 billion in revenue in the third quarter alone, up from $1.4 billion a year earlier. OpenAI, meanwhile, pulled back. Sources told the Financial Times that OpenAI cut its latest funding target to $20 billion from $30 billion and trimmed its expected valuation to $1.25 trillion from $1.4 trillion.  This is not the first time Son has publicly labeled a bet a mistake. After SoftBank’s WeWork wager soured in 2019, he said his investment judgment had been “really bad” and that he regretted it “in many ways.” WeWork never completed its planned IPO. The OpenAI commitment has already cost SoftBank on paper. Cryptopolitan reported in late 2025 that the company shed more than $100 billion in market value after selling out of Nvidia to concentrate on OpenAI, as traders began treating SoftBank stock as a proxy for Sam Altman’s firm. If you're reading this, you’re already ahead. Stay there with our newsletter.

Masayoshi Son calls Anthropic miss a 'mistake' a day before its IPO

SoftBank’s CEO Masayoshi Son has rued missing the chance to get in on Anthropic at an earlier stage on Wednesday, one day before the Claude maker debuts on the Nasdaq after its November 19 IPO.
Son, whose SoftBank is heavily invested in OpenAI, admitted that not having stakes in the AI lab hitting the market tomorrow is one of his investment “mistakes.” OpenAI, on the other hand, has pushed its own IPO into 2027, delaying Son and SoftBank’s cashout on their investment.
Did SoftBank invest in Anthropic?
Speaking at Nikkei’s “Dialogues on the Future” conference in Tokyo, SoftBank CEO Masayoshi Son talked about his firm’s all-in investment strategy in Sam Altman’s OpenAI, while skipping out on Dario Amodei’s Anthropic entirely.
Notably, Son admitted the difficulty of deciding which AI developer to hitch their wagon to in the AI race. His comments came as the OpenAI wagon that SoftBank jumped on continues to delay its IPO, while Anthropic, the company they have no stake in, is due to go public on Thursday under the ticker ANT on Nasdaq.
Despite the apparent lag, the SoftBank founder has not cooled on OpenAI, telling the forum that he measures success by how his firm’s investments create the future rather than by making money.
As of this report, SoftBank has committed up to $64.6 billion to OpenAI for a cumulative stake of roughly 13% held through Vision Fund 2. The final $10 billion tranche of a $30 billion follow-on investment in OpenAI was completed on October 1.
Who missed out on investing in Anthropic?
Son joins a list of publicly tracked big-name investors who will not hold any stakes in Anthropic when it debuts on the Nasdaq, either by completely passing up early investment rounds, walking away or losing their stakes before the IPO.
Sam Bankman-Fried (SBF) lands in the latter category. The disgraced FTX founder, who is currently serving a 25-year sentence at the FCI Terminal Island low-security federal prison in San Pedro, Los Angeles, had a 13.5% stake in Anthropic via a late 2021 $500 million investment.
However, the bankruptcy estate sold that Anthropic position for $1.3 billion in 2024 in a flash sale to repay FTX customers after the exchange collapsed in November 2022 due to an $8 billion hole in customer accounts.
Former FTX executives Caroline Ellison and Nishad Singh also forfeited stakes to the US government. Those stakes netted $627 million when they were sold. The proceeds from that sale were directed back to the FTX estate.
The other parties in the Anthropic stakes sale transactions are due for big paydays. One of them, Jane Street, will see the $1.34 billion it spent on FTX-linked Anthropic shares explode to be worth more than $35 billion upon stock listing.
OpenAI’s round shrinks as Anthropic races to list
The IPO timing lag is not the only point of contrast in the OpenAI versus Anthropic comparison. Anthropic priced its offering in a range of $70 to $78 a share, valuing the Claude maker at up to roughly $1.4 trillion, Fortune reported. The company booked $17 billion in revenue in the third quarter alone, up from $1.4 billion a year earlier.
OpenAI, meanwhile, pulled back. Sources told the Financial Times that OpenAI cut its latest funding target to $20 billion from $30 billion and trimmed its expected valuation to $1.25 trillion from $1.4 trillion.
This is not the first time Son has publicly labeled a bet a mistake. After SoftBank’s WeWork wager soured in 2019, he said his investment judgment had been “really bad” and that he regretted it “in many ways.” WeWork never completed its planned IPO.
The OpenAI commitment has already cost SoftBank on paper. Cryptopolitan reported in late 2025 that the company shed more than $100 billion in market value after selling out of Nvidia to concentrate on OpenAI, as traders began treating SoftBank stock as a proxy for Sam Altman’s firm.
If you're reading this, you’re already ahead. Stay there with our newsletter.
Article
World Bank explores USDC payments with Citi and CircleThe World Bank is examining whether USDC, a stablecoin pegged to the US dollar, can help to make payments for development quicker and cheaper, while still being able to monitor those funds easily. In its announcement made on November 17, the bank mentioned that it will cooperate with Circle and Citi Bank to determine how USDC payments can be executed through existing banking channels. An exploration, not a rollout The plan is still under consideration at the moment. Citi would be responsible for linking USDC transactions with existing banking networks. Circle would be responsible for handling USDC minting and settlement. The World Bank intends to engage only with borrowing countries where digital assets are allowed. No country has been named, nor has the launch date been revealed yet. President of the World Bank Group Ajay Banga said that there are several possible benefits, such as timely construction of clinics and timely payment to farmers before the harvest. What FundsChain already tracks FundsChain is the World Bank’s permissioned blockchain for tracking how development funds are spent. More than 350 projects in 96 countries are on its ledgers, but none is confirmed for the USDC proposal. The FundsChain FAQ says users can view transaction records based on their access permissions. The platform also uses a Traceability Index to measure how well transactions are supported by required documents. Missing records lower the score but do not delay disbursements. The questions the USDC leg raises Faster payments are not enough to prove that funds for development have gone to beneficiaries. The World Bank’s announcement is silent about how the proposed USDC system will verify recipients, confirm payments, keep records of audits, and protect the privacy of beneficiaries. It also leaves unanswered questions about how exchange rates and conversion fees would be documented. Those costs affect the amount of money recipients will receive in local currency. A BIS study reveals that fintech transfers are usually less expensive than bank wires. Stablecoin savings, meanwhile, could disappear after trading, conversion and withdrawal fees. Beyond costs, ECB President Christine Lagarde warned that widespread stablecoin use could weaken bank lending and make interest-rate policy less effective. World Bank, Citi and Circle Explore USDC Payment Rails Why Citi says the plumbing can work Citi has experimented with the connection of bank transfers and blockchain transactions. In their 2025 experiment, they collaborated with SWIFT to use test USDC on the Ethereum Sepolia network along with traditional money payments. The transfers were kept in sync through an escrow account, but they did not settle as part of the same transaction. Citi took another step in September 2026, when it reported live transactions on Swift’s blockchain ledger with First Abu Dhabi Bank and OCBC. That project was separate from the World Bank proposal. Looking ahead, Citi forecasts $1.9 trillion in global stablecoin issuance by 2030. The figure is a projection, not the current market size. A crowded push into bank stablecoins Major banks are also entering the stablecoin market. Reuters reported that 21 financial institutions, including Goldman Sachs, Bank of America and Citi, plan to launch a dollar-backed token in the first half of 2027. The competition is far from evenly matched. Reuters cited more than $180 billion in Tether tokens issued, compared with just $12.5 million of Société Générale’s dollar-backed token in circulation. Stablecoins are also gaining attention in development and humanitarian payments. The Circle Foundation is supporting separate digital-payment initiatives with UNDP and the World Food Programme. Meanwhile, Cryptopolitan previously reported on Citi’s collaboration with Coinbase for corporate payments. These initiatives show growing interest in stablecoin payments, although their benefits for development funding remain unproven. If you're reading this, you’re already ahead. Stay there with our newsletter.

World Bank explores USDC payments with Citi and Circle

The World Bank is examining whether USDC, a stablecoin pegged to the US dollar, can help to make payments for development quicker and cheaper, while still being able to monitor those funds easily.
In its announcement made on November 17, the bank mentioned that it will cooperate with Circle and Citi Bank to determine how USDC payments can be executed through existing banking channels.
An exploration, not a rollout
The plan is still under consideration at the moment. Citi would be responsible for linking USDC transactions with existing banking networks. Circle would be responsible for handling USDC minting and settlement.
The World Bank intends to engage only with borrowing countries where digital assets are allowed. No country has been named, nor has the launch date been revealed yet. President of the World Bank Group Ajay Banga said that there are several possible benefits, such as timely construction of clinics and timely payment to farmers before the harvest.
What FundsChain already tracks
FundsChain is the World Bank’s permissioned blockchain for tracking how development funds are spent. More than 350 projects in 96 countries are on its ledgers, but none is confirmed for the USDC proposal.
The FundsChain FAQ says users can view transaction records based on their access permissions. The platform also uses a Traceability Index to measure how well transactions are supported by required documents. Missing records lower the score but do not delay disbursements.
The questions the USDC leg raises
Faster payments are not enough to prove that funds for development have gone to beneficiaries. The World Bank’s announcement is silent about how the proposed USDC system will verify recipients, confirm payments, keep records of audits, and protect the privacy of beneficiaries. It also leaves unanswered questions about how exchange rates and conversion fees would be documented.
Those costs affect the amount of money recipients will receive in local currency. A BIS study reveals that fintech transfers are usually less expensive than bank wires.
Stablecoin savings, meanwhile, could disappear after trading, conversion and withdrawal fees. Beyond costs, ECB President Christine Lagarde warned that widespread stablecoin use could weaken bank lending and make interest-rate policy less effective.
World Bank, Citi and Circle Explore USDC Payment Rails
Why Citi says the plumbing can work
Citi has experimented with the connection of bank transfers and blockchain transactions. In their 2025 experiment, they collaborated with SWIFT to use test USDC on the Ethereum Sepolia network along with traditional money payments. The transfers were kept in sync through an escrow account, but they did not settle as part of the same transaction.
Citi took another step in September 2026, when it reported live transactions on Swift’s blockchain ledger with First Abu Dhabi Bank and OCBC. That project was separate from the World Bank proposal. Looking ahead, Citi forecasts $1.9 trillion in global stablecoin issuance by 2030. The figure is a projection, not the current market size.
A crowded push into bank stablecoins
Major banks are also entering the stablecoin market. Reuters reported that 21 financial institutions, including Goldman Sachs, Bank of America and Citi, plan to launch a dollar-backed token in the first half of 2027.
The competition is far from evenly matched. Reuters cited more than $180 billion in Tether tokens issued, compared with just $12.5 million of Société Générale’s dollar-backed token in circulation.
Stablecoins are also gaining attention in development and humanitarian payments. The Circle Foundation is supporting separate digital-payment initiatives with UNDP and the World Food Programme.
Meanwhile, Cryptopolitan previously reported on Citi’s collaboration with Coinbase for corporate payments. These initiatives show growing interest in stablecoin payments, although their benefits for development funding remain unproven.
If you're reading this, you’re already ahead. Stay there with our newsletter.
JPMorgan and Hyundai Card test AI agents that negotiate and make paymentsJPMorgan Chase and Hyundai Card are evaluating the extent to which AI agents could autonomously perform financial tasks, such as negotiating terms and processing payments without human intervention. The companies have just completed an initial proof of concept focused on a simulated foreign exchange transaction. A Hyundai Card AI agent made a foreign exchange request to increase euro liquidity in its Kinexys blockchain deposit account. This was the first-ever private blockchain experiment involving autonomous bank and customer agents in corporate treasuries, per JPMorgan. JPMorgan and Hyundai plan to run more tests in 2027 JPMorgan and Hyundai now intend to run a more complex, multi-agent AI testing phase in the first half of next year. Wee Kee Toh, who leads Kinexys Insights at Kinexys, noted that the next phase will test how autonomous agents communicate and engage with each other beyond executing transactions. Nevertheless, JPMorgan still has not announced a timeline for the possible commercialization of the technology, reflecting broader concerns among financial institutions about trusting autonomous AI systems with financial operations. Banks are still struggling to move AI agents beyond pilot projects, largely because of accountability, security, and compliance concerns. According to Toh, the organizations still need to develop appropriate governance models and ensure the implementation of the required safety measures. The use cases for autonomous financial agents are not limited to foreign exchange. Corporate treasury departments manage liquidity, currency exposure, and payment coordination, and ensure businesses have funds across multiple accounts. These processes often involve multiple institutions, making them complex. AI agents could eventually automate some of these tasks (e.g., liquidity requirements are identified, currency conversions requested, and payments between accounts coordinated). For multinational companies like Hyundai, this would help reduce delays in cross-border treasury operations. This would allow businesses to respond more quickly to changing financial needs — but transaction limits and verification mechanisms would be needed to ensure transfers aren’t made by accident. Developing technology wasn’t hard, he said — the real problem is accountability and risks of letting AI agents make financial transactions on their own. To date, Kinexys has created a model that restricts AI agents to transactional parameters and rejects any request that exceeds those boundaries. “You are going to draw a box around the agent. The agent will never be able to do anything out of the box,” Toh said. Last September, the bank highlighted some of its concerns regarding AI agents, noting that a command is just an intention and nothing more. Even if an agent is instructed not to access the internet, not to share data, or to stick strictly to a defined investment mandate, those guidelines only work if the tools, communication channels, and execution paths enforce them. The bank gave an example from a security evaluation in which an agent was explicitly told not to have any internet access but nevertheless managed to create malware that could run on real-world systems after accidentally discovering internet access during the evaluation. A larger review linked to that evaluation also found unauthorized access affecting numerous organizations. In another set of evaluations, agents co-opted an internal package service to act as a persistent message board, passed discoveries across separate runs, and compromised parts of external production infrastructure. They also bypassed internal research infrastructure and accessed sensitive secrets on a massive scale. JPMorgan Chase has been working on AI agents since June JPMorgan Chase initially shared its plans to deploy long-running agents in June. In a whitepaper published that same month, the firm highlighted a specific treasury operations scenario where an AI agent managing a payment exceeding $1 million detected a change in the transaction’s currency. That one transaction reportedly averted a possible $2.3 million currency mismatch. Later in the year, the bank took a further step by implementing defensive measures to protect customer accounts from external agents attempting to access them for liability purposes. Speaking during the October earnings call, JPMorgan’s Chief Executive Officer Jamie Dimon clarified that the bank excludes outside agents because of liability issues associated with customer accounts. He asserted customers themselves assume that liability when deciding to grant an agent access to their accounts. That same month, the bank partnered with Mirakl on agentic commerce, which helps with in-house payment processing. If you're reading this, you’re already ahead. Stay there with our newsletter.

JPMorgan and Hyundai Card test AI agents that negotiate and make payments

JPMorgan Chase and Hyundai Card are evaluating the extent to which AI agents could autonomously perform financial tasks, such as negotiating terms and processing payments without human intervention.
The companies have just completed an initial proof of concept focused on a simulated foreign exchange transaction. A Hyundai Card AI agent made a foreign exchange request to increase euro liquidity in its Kinexys blockchain deposit account. This was the first-ever private blockchain experiment involving autonomous bank and customer agents in corporate treasuries, per JPMorgan.
JPMorgan and Hyundai plan to run more tests in 2027
JPMorgan and Hyundai now intend to run a more complex, multi-agent AI testing phase in the first half of next year. Wee Kee Toh, who leads Kinexys Insights at Kinexys, noted that the next phase will test how autonomous agents communicate and engage with each other beyond executing transactions.
Nevertheless, JPMorgan still has not announced a timeline for the possible commercialization of the technology, reflecting broader concerns among financial institutions about trusting autonomous AI systems with financial operations. Banks are still struggling to move AI agents beyond pilot projects, largely because of accountability, security, and compliance concerns.
According to Toh, the organizations still need to develop appropriate governance models and ensure the implementation of the required safety measures.
The use cases for autonomous financial agents are not limited to foreign exchange. Corporate treasury departments manage liquidity, currency exposure, and payment coordination, and ensure businesses have funds across multiple accounts. These processes often involve multiple institutions, making them complex.
AI agents could eventually automate some of these tasks (e.g., liquidity requirements are identified, currency conversions requested, and payments between accounts coordinated). For multinational companies like Hyundai, this would help reduce delays in cross-border treasury operations.
This would allow businesses to respond more quickly to changing financial needs — but transaction limits and verification mechanisms would be needed to ensure transfers aren’t made by accident.
Developing technology wasn’t hard, he said — the real problem is accountability and risks of letting AI agents make financial transactions on their own.
To date, Kinexys has created a model that restricts AI agents to transactional parameters and rejects any request that exceeds those boundaries. “You are going to draw a box around the agent. The agent will never be able to do anything out of the box,” Toh said.
Last September, the bank highlighted some of its concerns regarding AI agents, noting that a command is just an intention and nothing more. Even if an agent is instructed not to access the internet, not to share data, or to stick strictly to a defined investment mandate, those guidelines only work if the tools, communication channels, and execution paths enforce them.
The bank gave an example from a security evaluation in which an agent was explicitly told not to have any internet access but nevertheless managed to create malware that could run on real-world systems after accidentally discovering internet access during the evaluation. A larger review linked to that evaluation also found unauthorized access affecting numerous organizations.
In another set of evaluations, agents co-opted an internal package service to act as a persistent message board, passed discoveries across separate runs, and compromised parts of external production infrastructure. They also bypassed internal research infrastructure and accessed sensitive secrets on a massive scale.
JPMorgan Chase has been working on AI agents since June
JPMorgan Chase initially shared its plans to deploy long-running agents in June. In a whitepaper published that same month, the firm highlighted a specific treasury operations scenario where an AI agent managing a payment exceeding $1 million detected a change in the transaction’s currency. That one transaction reportedly averted a possible $2.3 million currency mismatch.
Later in the year, the bank took a further step by implementing defensive measures to protect customer accounts from external agents attempting to access them for liability purposes.
Speaking during the October earnings call, JPMorgan’s Chief Executive Officer Jamie Dimon clarified that the bank excludes outside agents because of liability issues associated with customer accounts.
He asserted customers themselves assume that liability when deciding to grant an agent access to their accounts. That same month, the bank partnered with Mirakl on agentic commerce, which helps with in-house payment processing.
If you're reading this, you’re already ahead. Stay there with our newsletter.
Article
FinCEN proposes tighter ID checks for business owners using ITINsOn November 17, 2026, the Financial Crimes Enforcement Network (FinCEN) proposed new rules for verifying who owns or controls a business. Under the proposal, banks could no longer accept Individual Taxpayer Identification Numbers (ITINs) to identify beneficial owners. The change could make banking more difficult for some foreign-owned firms. However, this measure wouldn’t prohibit ITIN holders from opening an account or make a difference in the Employer Identification Number of the businesses. According to FinCEN, the initiative will prevent financial crimes. Community bankers, on the other hand, say that the tough measures may lead to some unnecessary hindrances for legitimate customers. What an ITIN proves, and what it doesn’t An ITIN refers to a nine-digit number issued by the Internal Revenue Service (IRS) for federal tax purposes to people who are not eligible for a Social Security number. The main purpose of an ITIN is taxation. According to the IRS, an ITIN does not give a person the right to work, provide information about immigration status, or be used as identification beyond the federal tax system. FinCEN proposes ITIN restrictions for business owners at US banks A tighter screen on who owns a business The ITIN restriction proposed by FinCEN would amend the Customer Due Diligence (CDD) Rule. This rule requires banks and other covered financial institutions to identify and verify their business customers’ beneficial owners. A beneficial owner is someone who owns at least 25% of a company or has significant responsibility for managing it. The proposal is made following an executive order from May 19, 2026. Banks could no longer accept ITINs as identifying numbers for beneficial owners. Non-U.S. owners could still provide passports or other specified government documents. As a measure of precaution, banks would be entitled to investigate someone’s immigration status or work permit if they feel there may be certain risks involved. These restrictions apply only to some customers of the bank. Relief from repeated ownership checks FinCEN is pushing for stricter identification rules while easing repeat checks. Specifically, FinCEN notes that banks are no longer obligated to verify a company’s owner whenever the company opens additional accounts, as required under an order issued by FinCEN on February 13. Verification is necessary when opening accounts for the first time, when earlier information becomes unreliable, or when risks deem it necessary. The November proposal would make that relief part of the CDD Rule. Separately, an August 11 final rule exempted U.S. companies and persons from Corporate Transparency Act reporting. Certain foreign companies must still file. The August rule concerns government reporting, not bank verification. Community bankers warn of over-screening Community bankers worry the proposed ITIN restriction could go too far. In a June 23 letter, the Independent Community Bankers of America warned that treating ITIN use alone as suspicious could delay mortgages, farm loans and small-business credit. It urged regulators not to make banks “legal-status gatekeepers.” FinCEN’s June advisory identifies certain ITIN-related practices as potential warning signs, not proof of wrongdoing. Banks must distinguish suspicious activity from routine tax identification. Immigrant businesses and the market-access question An IZA study counted 3.9 million active immigrant business owners in 2024, up from 3.0 million before the pandemic. Growth included construction, transportation and professional services. The figures do not measure ITIN use. Still, stricter checks could complicate U.S. banking and cross-border transactions for foreign entrepreneurs. The Financial Action Task Force advocates safeguards that preserve financial inclusion. Meanwhile, Cryptopolitan reported $12.7 billion in suspected crypto-scam activity flagged in 2023–2025 filings. The figure represents neither confirmed losses nor evidence of ITIN misuse. FinCEN must balance crime prevention with banking access for legitimate businesses. Comments close 60 days after Federal Register publication. The smartest crypto minds already read our newsletter. Want in? Join them.

FinCEN proposes tighter ID checks for business owners using ITINs

On November 17, 2026, the Financial Crimes Enforcement Network (FinCEN) proposed new rules for verifying who owns or controls a business. Under the proposal, banks could no longer accept Individual Taxpayer Identification Numbers (ITINs) to identify beneficial owners. The change could make banking more difficult for some foreign-owned firms.
However, this measure wouldn’t prohibit ITIN holders from opening an account or make a difference in the Employer Identification Number of the businesses. According to FinCEN, the initiative will prevent financial crimes. Community bankers, on the other hand, say that the tough measures may lead to some unnecessary hindrances for legitimate customers.
What an ITIN proves, and what it doesn’t
An ITIN refers to a nine-digit number issued by the Internal Revenue Service (IRS) for federal tax purposes to people who are not eligible for a Social Security number.
The main purpose of an ITIN is taxation. According to the IRS, an ITIN does not give a person the right to work, provide information about immigration status, or be used as identification beyond the federal tax system.
FinCEN proposes ITIN restrictions for business owners at US banks
A tighter screen on who owns a business
The ITIN restriction proposed by FinCEN would amend the Customer Due Diligence (CDD) Rule. This rule requires banks and other covered financial institutions to identify and verify their business customers’ beneficial owners. A beneficial owner is someone who owns at least 25% of a company or has significant responsibility for managing it.
The proposal is made following an executive order from May 19, 2026. Banks could no longer accept ITINs as identifying numbers for beneficial owners. Non-U.S. owners could still provide passports or other specified government documents.
As a measure of precaution, banks would be entitled to investigate someone’s immigration status or work permit if they feel there may be certain risks involved. These restrictions apply only to some customers of the bank.
Relief from repeated ownership checks
FinCEN is pushing for stricter identification rules while easing repeat checks. Specifically, FinCEN notes that banks are no longer obligated to verify a company’s owner whenever the company opens additional accounts, as required under an order issued by FinCEN on February 13. Verification is necessary when opening accounts for the first time, when earlier information becomes unreliable, or when risks deem it necessary.
The November proposal would make that relief part of the CDD Rule. Separately, an August 11 final rule exempted U.S. companies and persons from Corporate Transparency Act reporting. Certain foreign companies must still file. The August rule concerns government reporting, not bank verification.
Community bankers warn of over-screening
Community bankers worry the proposed ITIN restriction could go too far. In a June 23 letter, the Independent Community Bankers of America warned that treating ITIN use alone as suspicious could delay mortgages, farm loans and small-business credit. It urged regulators not to make banks “legal-status gatekeepers.”
FinCEN’s June advisory identifies certain ITIN-related practices as potential warning signs, not proof of wrongdoing. Banks must distinguish suspicious activity from routine tax identification.
Immigrant businesses and the market-access question
An IZA study counted 3.9 million active immigrant business owners in 2024, up from 3.0 million before the pandemic. Growth included construction, transportation and professional services. The figures do not measure ITIN use.
Still, stricter checks could complicate U.S. banking and cross-border transactions for foreign entrepreneurs. The Financial Action Task Force advocates safeguards that preserve financial inclusion.
Meanwhile, Cryptopolitan reported $12.7 billion in suspected crypto-scam activity flagged in 2023–2025 filings. The figure represents neither confirmed losses nor evidence of ITIN misuse. FinCEN must balance crime prevention with banking access for legitimate businesses. Comments close 60 days after Federal Register publication.
The smartest crypto minds already read our newsletter. Want in? Join them.
Coinbase Says AI Stopped $4.7M in Deepfake Musk ‘Tesla Giveaway’ Scam Coinbase has said that its AI-powered scam defenses blocked $4.7 million from reaching attackers who used a deepfake of Elon Musk and hijacked the YouTube livestream of Tesla’s “flying car” demo. More than 500,000 people tuned in to the fake livestream where they were lured to send crypto via a QR code on their screen with the promise of it getting doubled.  Coinbase reported that within about two hours there were hundreds of Coinbase users who had tried to send millions of dollars to the attackers wallets. The exchange’s automated defenses caught roughly 91% of it without a single employee stepping in.  Scammers used an AI deepfake of Elon Musk to promote a fake Tesla crypto giveaway to a livestream audience of more than 500,000 people. In about two hours, hundreds of Coinbase customers tried to send millions to the scammers. Our AI-powered scam defenses stopped $4.7 million… — Coinbase Support (@CoinbaseSupport) November 17, 2026 2026 Has Already Broken the Record for Crypto Theft This year has been an abysmal year for crypto related hacks and scams. Data from DeFiLlama shows that there have been 343 crypto thefts year to date with the amount being stolen $8.78 billion. October’s Binance hack makes up most of that and investigators have tied the largest crypto theft in history to TraderTraitor, a North Korean hacking group.  While several hacks throughout the year have grabbed the headlines, scams are actually growing faster. TRM Labs said in August that reported losses to deepfake scams in 2026 had already surpassed the total seen in 2025 by 263%. These sort of scams don’t need to break into anything. They convince people to move their money and to the exchange processing it.  Coinbase Fought the Same Scam by Hand in 2020 The fact is the exchange has actually come across a very similar double your money type of scam a couple of years ago. Back in July 2020, a bunch of well known twitter accounts were hacked, with Musk being one, and an almost similar offer was posted then. Coinbase’s response back then was to manually blacklist the Bitcoin address the scammers were using. That fix only kicked in once the address was public and money was already flowing. Fast forward six years and the core gist of the scam remains the same. It’s just the method of doing it has evolved from a hijacked tweet to a hyper realistic Musk on video nudging viewers to a QR code, and that kind of content can reach half a million people in a single stream. Over the span of the last six years, the defense has changed too. Coinbase’s software effectively decided its own customers were about to make a costly mistake and stepped in before the money left. Exchanges have long treated user error as the user’s problem. With deepfakes getting cheaper and more convincing every month, Coinbase is now treating it as something its security stack has to catch.  The smartest crypto minds already read our newsletter. Want in? Join them.

Coinbase Says AI Stopped $4.7M in Deepfake Musk ‘Tesla Giveaway’ Scam 

Coinbase has said that its AI-powered scam defenses blocked $4.7 million from reaching attackers who used a deepfake of Elon Musk and hijacked the YouTube livestream of Tesla’s “flying car” demo. More than 500,000 people tuned in to the fake livestream where they were lured to send crypto via a QR code on their screen with the promise of it getting doubled.
Coinbase reported that within about two hours there were hundreds of Coinbase users who had tried to send millions of dollars to the attackers wallets. The exchange’s automated defenses caught roughly 91% of it without a single employee stepping in.
Scammers used an AI deepfake of Elon Musk to promote a fake Tesla crypto giveaway to a livestream audience of more than 500,000 people.
In about two hours, hundreds of Coinbase customers tried to send millions to the scammers. Our AI-powered scam defenses stopped $4.7 million…
— Coinbase Support (@CoinbaseSupport) November 17, 2026
2026 Has Already Broken the Record for Crypto Theft
This year has been an abysmal year for crypto related hacks and scams. Data from DeFiLlama shows that there have been 343 crypto thefts year to date with the amount being stolen $8.78 billion. October’s Binance hack makes up most of that and investigators have tied the largest crypto theft in history to TraderTraitor, a North Korean hacking group.
While several hacks throughout the year have grabbed the headlines, scams are actually growing faster. TRM Labs said in August that reported losses to deepfake scams in 2026 had already surpassed the total seen in 2025 by 263%. These sort of scams don’t need to break into anything. They convince people to move their money and to the exchange processing it.
Coinbase Fought the Same Scam by Hand in 2020
The fact is the exchange has actually come across a very similar double your money type of scam a couple of years ago. Back in July 2020, a bunch of well known twitter accounts were hacked, with Musk being one, and an almost similar offer was posted then. Coinbase’s response back then was to manually blacklist the Bitcoin address the scammers were using. That fix only kicked in once the address was public and money was already flowing.
Fast forward six years and the core gist of the scam remains the same. It’s just the method of doing it has evolved from a hijacked tweet to a hyper realistic Musk on video nudging viewers to a QR code, and that kind of content can reach half a million people in a single stream.
Over the span of the last six years, the defense has changed too. Coinbase’s software effectively decided its own customers were about to make a costly mistake and stepped in before the money left. Exchanges have long treated user error as the user’s problem. With deepfakes getting cheaper and more convincing every month, Coinbase is now treating it as something its security stack has to catch.
The smartest crypto minds already read our newsletter. Want in? Join them.
Article
Equinor battles ongoing cyberattack as Norway remains on high alertEquinor, an international energy company headquartered in Norway, said on November 17, 2026, that some of its IT systems were under cyberattack. Oil and gas production remains normal, and the company has seen no signs of stolen data. The incident has received significant attention because Norway is a major supplier of gas for Europe. What Equinor has confirmed, and what it hasn’t In its November 17 statement, Equinor described the cyberattack as ongoing. The company said it had mobilised resources to stop the attack and was notifying relevant authorities. Equinor also stated: There are no indications that data has been stolen. The attack has not affected Equinor’s operations and production. The statement did not mention who the attackers were, how they got in, or whether a malware was involved. Equinor also did not say anything about whether its industrial control systems had been affected in any way. While the company said that nothing indicated any data theft, the initial assessment did not exclude the possibility of it happening. Equinor IT cyberattack: Norway gas production unaffected Where an IT breach can still reach the machinery Equinor confirmed an attack on its information technology (IT) systems, which handle email, business software, and company data. However, it has not said whether the attackers reached the systems controlling its machinery. Those systems fall under operational technology (OT), which includes sensors, controllers, and engineering systems. An attack on IT systems does not automatically give attackers control of OT systems. IT and OT systems have different purposes, but that does not mean their networks are entirely separate from each other. Shared login credentials, contractor connections, and weak network security can give attackers a way to move between them. A 2026 Dragos assessment indicate that 29% of companies within the oil and gas sector were severely lacking in IT/OT separation. A broader cyberattacks study by Verizon in 2026 shows that vulnerabilities in software were used to obtain initial access in 31% of attacks. Neither of the reports detailed what transpired at Equinor. An assault on IT systems does not, in itself, indicate that operations were disrupted or that the attackers were able to seize control of any machines. Such developments would require different proof. Norway on its highest openly declared cyber alert Norway was already on high cyber alert when Equinor disclosed the attack. On October 27, the National Security Authority (NSM) raised the country’s cyber-risk level to HIGH, the second-highest of five levels. The agency made its assessment public for the first time. It also reported that successful cyber operations against Norwegian organisations had tripled over six months. Aftenposten reported other incidents involving data stolen from Gripr, leaked SAS passenger records through Amadeus, and manipulated refrigeration temperatures at an Enebakk warehouse. No evidence publicly links these incidents to Equinor or one another. Why a Norwegian outage would be a European problem The consequences could extend beyond Norway if a cyberattack disrupted gas production. Norway is the world’s fourth-largest gas exporter. Its 2025 exports equalled more than 30% of EU and UK gas consumption. About 95% of Norwegian gas travels through subsea pipelines. The Norwegian state also owns 67% of Equinor. No supply disruption has been reported. However, the IEA warned that Europe entered the 2026–27 winter with unusually low gas storage. A prolonged outage could tighten supplies and drive prices higher. That remains a potential risk, not an observed consequence of Equinor’s cyberattack. Echoes of Colonial Pipeline The Colonial Pipeline incident illustrates how an attack in a company’s IT system can affect the delivery of oil and gas without the interference of industrial controls. After a ransomware attack in 2021, the operator of the pipeline stopped its operations temporarily as a precaution. As reported by Cryptopolitan, Colonial paid $4.4 million in Bitcoin. U.S. officials later recovered $2.3 million of that amount. Cybersecurity reports suggest industrial companies continue to face considerable pressure. Dragos recorded 1,266 industrial ransomware incidents in Q3 2026, up 11% from the previous quarter. However, it found no cases of ransomware operators directly manipulating industrial controls. The report from Zscaler indicated an increase of more than 275% in data theft due to ransomware between April 2025 and March 2026. Neither report provides any information on what transpired in the case involving Equinor. It still needs to be determined how widespread the attack was. The smartest crypto minds already read our newsletter. Want in? Join them.

Equinor battles ongoing cyberattack as Norway remains on high alert

Equinor, an international energy company headquartered in Norway, said on November 17, 2026, that some of its IT systems were under cyberattack. Oil and gas production remains normal, and the company has seen no signs of stolen data. The incident has received significant attention because Norway is a major supplier of gas for Europe.
What Equinor has confirmed, and what it hasn’t
In its November 17 statement, Equinor described the cyberattack as ongoing. The company said it had mobilised resources to stop the attack and was notifying relevant authorities. Equinor also stated:
There are no indications that data has been stolen. The attack has not affected Equinor’s operations and production.
The statement did not mention who the attackers were, how they got in, or whether a malware was involved. Equinor also did not say anything about whether its industrial control systems had been affected in any way. While the company said that nothing indicated any data theft, the initial assessment did not exclude the possibility of it happening.
Equinor IT cyberattack: Norway gas production unaffected
Where an IT breach can still reach the machinery
Equinor confirmed an attack on its information technology (IT) systems, which handle email, business software, and company data. However, it has not said whether the attackers reached the systems controlling its machinery. Those systems fall under operational technology (OT), which includes sensors, controllers, and engineering systems. An attack on IT systems does not automatically give attackers control of OT systems.
IT and OT systems have different purposes, but that does not mean their networks are entirely separate from each other. Shared login credentials, contractor connections, and weak network security can give attackers a way to move between them. A 2026 Dragos assessment indicate that 29% of companies within the oil and gas sector were severely lacking in IT/OT separation. A broader cyberattacks study by Verizon in 2026 shows that vulnerabilities in software were used to obtain initial access in 31% of attacks.
Neither of the reports detailed what transpired at Equinor. An assault on IT systems does not, in itself, indicate that operations were disrupted or that the attackers were able to seize control of any machines. Such developments would require different proof.
Norway on its highest openly declared cyber alert
Norway was already on high cyber alert when Equinor disclosed the attack. On October 27, the National Security Authority (NSM) raised the country’s cyber-risk level to HIGH, the second-highest of five levels. The agency made its assessment public for the first time. It also reported that successful cyber operations against Norwegian organisations had tripled over six months.
Aftenposten reported other incidents involving data stolen from Gripr, leaked SAS passenger records through Amadeus, and manipulated refrigeration temperatures at an Enebakk warehouse. No evidence publicly links these incidents to Equinor or one another.
Why a Norwegian outage would be a European problem
The consequences could extend beyond Norway if a cyberattack disrupted gas production. Norway is the world’s fourth-largest gas exporter. Its 2025 exports equalled more than 30% of EU and UK gas consumption. About 95% of Norwegian gas travels through subsea pipelines. The Norwegian state also owns 67% of Equinor.
No supply disruption has been reported. However, the IEA warned that Europe entered the 2026–27 winter with unusually low gas storage. A prolonged outage could tighten supplies and drive prices higher. That remains a potential risk, not an observed consequence of Equinor’s cyberattack.
Echoes of Colonial Pipeline
The Colonial Pipeline incident illustrates how an attack in a company’s IT system can affect the delivery of oil and gas without the interference of industrial controls. After a ransomware attack in 2021, the operator of the pipeline stopped its operations temporarily as a precaution. As reported by Cryptopolitan, Colonial paid $4.4 million in Bitcoin. U.S. officials later recovered $2.3 million of that amount.
Cybersecurity reports suggest industrial companies continue to face considerable pressure. Dragos recorded 1,266 industrial ransomware incidents in Q3 2026, up 11% from the previous quarter. However, it found no cases of ransomware operators directly manipulating industrial controls.
The report from Zscaler indicated an increase of more than 275% in data theft due to ransomware between April 2025 and March 2026. Neither report provides any information on what transpired in the case involving Equinor. It still needs to be determined how widespread the attack was.
The smartest crypto minds already read our newsletter. Want in? Join them.
Waymo suspends all US robotaxi freeway rides after California banWaymo has stopped offering driverless taxi rides on highways across the US after California banned its self-driving cars from using highways in the state. California’s Department of Motor Vehicles (DMV) said the cars were an “imminent hazard to public safety.” Waymo, owned by Google’s parent company Alphabet, then paused highway rides nationwide to review its self-driving software. California pulled Waymo off its highways California’s Department of Motor Vehicles (DMV) banned Waymo’s driverless cars from using highways on November 11. It announced the ban on November 17, according to Reuters. The agency said Waymo’s driverless highway service “poses an imminent hazard to public safety.” The ban came after several Waymo cars drove into areas closed for roadwork. The DMV said three Waymo cars entered a closed part of U.S. Route 101 near Burlingame while roadwork was taking place. Another incident took place on Interstate 280 in San Francisco. A Waymo car tried to pass two California Highway Patrol cars that were slowing traffic to keep drivers safe. According to the San Francisco Chronicle, a police officer had to block the Waymo car to stop it from passing. The car then moved into another lane and slowed down. The San Francisco Business Times also reported that the National Highway Traffic Safety Administration had asked Waymo to explain what had happened. The DMV claimed Waymo had failed to fix repeated problems with its cars driving into roadwork areas, even after earlier restrictions and efforts to solve the issue. The ban will stay in place until Waymo proves it has fixed the problems. Nationwide pause not a part of order California’s ban only applies to the state, however, Waymo has stopped highway rides throughout the US. The company said it is doing this to check its self-driving technology. Waymo offers driverless taxi rides in 15 US cities, but highway rides were only available in a few, including San Francisco and Los Angeles. Passengers can still use Waymo on regular roads. The company added that it provides more than 500,000 rides each week. The change has affected passengers in the San Francisco Bay Area, especially those traveling between San Francisco and Silicon Valley. Waymo cars now have to avoid highways, making some journeys longer. Waymo spokesperson Dana Gallagher told TechCrunch that “safety is at the foundation of everything we do.” Gallagher added that no one was injured and no crashes occurred during the incidents on U.S. Route 101. Waymo’s problems with road construction zones Waymo has had similar problems before. The company said seven of its cars entered highway lanes closed for roadwork in the San Francisco Bay Area on May 18, 2026. The incidents were disclosed in a recall report filed with the National Highway Traffic Safety Administration (NHTSA) on June 17. The cars passed between traffic cones marking the closed lanes. The report said their software either failed to notice the roadwork or tried to avoid other dangers on the road. The recall covered 3,871 fifth-generation self-driving systems approved for driverless highway trips. Waymo also faced similar problems in Arizona. The report listed six incidents in Phoenix in April, when its cars drove past signs showing that highway ramps were closed and entered areas marked for construction. According to International Business Times, published by Yahoo, Waymo cut back on highway trips after the May incidents. It brought the service back in late July after updating its software to better recognize roadwork areas. However, regulators said the latest incidents show that the updates have not solved the problem. Setback as Tesla and Zoox close in The suspension comes as Waymo is trying to grow its driverless taxi business. In November 2025, the company began offering highway rides to passengers in the San Francisco Bay Area, Phoenix and Los Angeles. This was meant to make travel between cities, suburbs and airports faster. The suspension could potentially slow Waymo’s growth as Tesla and Amazon’s Zoox also expand their driverless taxi services. Waymo has not said when highway rides will commence once again. The company said it is “working with the California DMV and other state officials” to restart the service. The smartest crypto minds already read our newsletter. Want in? Join them.

Waymo suspends all US robotaxi freeway rides after California ban

Waymo has stopped offering driverless taxi rides on highways across the US after California banned its self-driving cars from using highways in the state. California’s Department of Motor Vehicles (DMV) said the cars were an “imminent hazard to public safety.”
Waymo, owned by Google’s parent company Alphabet, then paused highway rides nationwide to review its self-driving software.
California pulled Waymo off its highways
California’s Department of Motor Vehicles (DMV) banned Waymo’s driverless cars from using highways on November 11. It announced the ban on November 17, according to Reuters. The agency said Waymo’s driverless highway service “poses an imminent hazard to public safety.”
The ban came after several Waymo cars drove into areas closed for roadwork. The DMV said three Waymo cars entered a closed part of U.S. Route 101 near Burlingame while roadwork was taking place.
Another incident took place on Interstate 280 in San Francisco. A Waymo car tried to pass two California Highway Patrol cars that were slowing traffic to keep drivers safe. According to the San Francisco Chronicle, a police officer had to block the Waymo car to stop it from passing. The car then moved into another lane and slowed down.
The San Francisco Business Times also reported that the National Highway Traffic Safety Administration had asked Waymo to explain what had happened.
The DMV claimed Waymo had failed to fix repeated problems with its cars driving into roadwork areas, even after earlier restrictions and efforts to solve the issue.
The ban will stay in place until Waymo proves it has fixed the problems.
Nationwide pause not a part of order
California’s ban only applies to the state, however, Waymo has stopped highway rides throughout the US. The company said it is doing this to check its self-driving technology.
Waymo offers driverless taxi rides in 15 US cities, but highway rides were only available in a few, including San Francisco and Los Angeles. Passengers can still use Waymo on regular roads. The company added that it provides more than 500,000 rides each week.
The change has affected passengers in the San Francisco Bay Area, especially those traveling between San Francisco and Silicon Valley. Waymo cars now have to avoid highways, making some journeys longer.
Waymo spokesperson Dana Gallagher told TechCrunch that “safety is at the foundation of everything we do.” Gallagher added that no one was injured and no crashes occurred during the incidents on U.S. Route 101.
Waymo’s problems with road construction zones
Waymo has had similar problems before. The company said seven of its cars entered highway lanes closed for roadwork in the San Francisco Bay Area on May 18, 2026. The incidents were disclosed in a recall report filed with the National Highway Traffic Safety Administration (NHTSA) on June 17.
The cars passed between traffic cones marking the closed lanes. The report said their software either failed to notice the roadwork or tried to avoid other dangers on the road. The recall covered 3,871 fifth-generation self-driving systems approved for driverless highway trips.
Waymo also faced similar problems in Arizona. The report listed six incidents in Phoenix in April, when its cars drove past signs showing that highway ramps were closed and entered areas marked for construction.
According to International Business Times, published by Yahoo, Waymo cut back on highway trips after the May incidents. It brought the service back in late July after updating its software to better recognize roadwork areas.
However, regulators said the latest incidents show that the updates have not solved the problem.
Setback as Tesla and Zoox close in
The suspension comes as Waymo is trying to grow its driverless taxi business. In November 2025, the company began offering highway rides to passengers in the San Francisco Bay Area, Phoenix and Los Angeles. This was meant to make travel between cities, suburbs and airports faster.
The suspension could potentially slow Waymo’s growth as Tesla and Amazon’s Zoox also expand their driverless taxi services.
Waymo has not said when highway rides will commence once again. The company said it is “working with the California DMV and other state officials” to restart the service.
The smartest crypto minds already read our newsletter. Want in? Join them.
AMZN+0.25%
GOOGL+0.28%
GOOGUS+1.32%
Article
AI models flagged civilians as targets in Vals war simulationsAI-testing company Vals reported on November 17 that many leading AI models indicated civilians, ambulances, and a school as potential military targets during image-based tests. Some models initially refused the requests but complied when researchers presented the exercise as a game. These results show that existing safeguards may not be enough to prevent misuse of these AI models. However, Vals did not test how these models would perform in actual weapons systems. ‘Must not overthink safety’ Vals has evaluated the AI models using drone scenes, battlefield surveys, and satellite images. In the drone scenes, GPT-6 Astra managed to identify at least one target in around 6% of its interactions with standard requests. However, when positioned within a gaming context, this number increased dramatically to 92%. GPT-6.1 Sol responded similarly, i.e. 17% in the first case and 90% in the second. Kimi K3.1 Pro, shown in a published reasoning trace, seems to disable safety concerns before responding. Vals also managed to establish that no safety classifier stopped its requests although some models refused its requests on their own. The school next to the military complex Vals also tested AI models using satellite images of Minab, Iran, taken before a February 2026 strike. The images showed an elementary school beside a military complex. Claude Opus 5.5 identified the school as a target in 69% of standard trials. That rose to 72% when researchers presented the exercise as a game. GPT-6.1 Sol’s rate increased from 0% to 41%. Each condition involved 32 trials. The school was hit in February 2026. United Nations experts later reported that at least 168 were dead. But Vals didn’t investigate the attack or claim that its tests provided any answers regarding what happened. AI Models Flag Civilians in Vals War Simulations A chatbot is not a weapon system Finding a potential target does not imply that an AI model is able to decide to carry out an attack. Vals only investigated how models give information for potential targets. The researchers did not investigate whether the results were correct or in line with the laws of war. The Council on Foreign Relations has pointed out that software that guides targets doesn’t constitute an autonomous weapon in itself. Legal researchers note that identifying targets via AI still needs confirmation. Meanwhile, the Brennan Center states that human monitoring cannot ensure that mistakes won’t threaten civilians’ safety. The fixes on the table Vals wants AI developers to strengthen safeguards and train models to resist misuse involving conventional weapons. The concern extends beyond its findings. Anthropic’s research also found that AI models could assist with tactical intelligence work. Anthropic says it has introduced classifiers to detect and block misuse. Governments have a role as well. SIPRI recommends setting clear requirements for military AI suppliers. It also urges governments to test suppliers’ claims independently and establish accountability before awarding contracts. A growing market and a Pentagon feud Military AI is also attracting more investment. Grand View Research estimates the market at $11.7 billion in 2026 and expects it to reach $19.3 billion by 2030. As Cryptopolitan reported, the Pentagon aims to establish AutoWarCom in October 2027. However, the Pentagon’s partnerships with AI companies have also sparked disputes. A congressional report says Anthropic, Google, OpenAI, and xAI received contracts in 2025 worth up to $200 million each. These amounts are spending limits, not guaranteed payments. Anthropic opposed using its technology for mass domestic surveillance and fully autonomous weapons. The report also notes that the Pentagon is not publicly known to use any frontier AI model in an autonomous weapon. Geneva splits over binding rules Countries are also divided over how to regulate autonomous weapons. At UN talks in Geneva, most of the 128 states party to the Convention on Certain Conventional Weapons support negotiations on binding rules, Reuters reported. The US, Russia and Israel oppose that approach. They argue that existing humanitarian law already covers autonomous weapons. Russia also delayed the opening discussions over whether observers could speak. Washington proposed five years of negotiations toward a non-binding agreement. The meeting’s chair suggested a two-year process to develop a binding protocol. Meanwhile, UN Secretary-General António Guterres urged governments to agree on clear prohibitions and restrictions. The smartest crypto minds already read our newsletter. Want in? Join them.

AI models flagged civilians as targets in Vals war simulations

AI-testing company Vals reported on November 17 that many leading AI models indicated civilians, ambulances, and a school as potential military targets during image-based tests. Some models initially refused the requests but complied when researchers presented the exercise as a game. These results show that existing safeguards may not be enough to prevent misuse of these AI models. However, Vals did not test how these models would perform in actual weapons systems.
‘Must not overthink safety’
Vals has evaluated the AI models using drone scenes, battlefield surveys, and satellite images. In the drone scenes, GPT-6 Astra managed to identify at least one target in around 6% of its interactions with standard requests. However, when positioned within a gaming context, this number increased dramatically to 92%. GPT-6.1 Sol responded similarly, i.e. 17% in the first case and 90% in the second. Kimi K3.1 Pro, shown in a published reasoning trace, seems to disable safety concerns before responding. Vals also managed to establish that no safety classifier stopped its requests although some models refused its requests on their own.
The school next to the military complex
Vals also tested AI models using satellite images of Minab, Iran, taken before a February 2026 strike. The images showed an elementary school beside a military complex. Claude Opus 5.5 identified the school as a target in 69% of standard trials. That rose to 72% when researchers presented the exercise as a game. GPT-6.1 Sol’s rate increased from 0% to 41%. Each condition involved 32 trials.
The school was hit in February 2026. United Nations experts later reported that at least 168 were dead. But Vals didn’t investigate the attack or claim that its tests provided any answers regarding what happened.
AI Models Flag Civilians in Vals War Simulations
A chatbot is not a weapon system
Finding a potential target does not imply that an AI model is able to decide to carry out an attack. Vals only investigated how models give information for potential targets. The researchers did not investigate whether the results were correct or in line with the laws of war.
The Council on Foreign Relations has pointed out that software that guides targets doesn’t constitute an autonomous weapon in itself. Legal researchers note that identifying targets via AI still needs confirmation. Meanwhile, the Brennan Center states that human monitoring cannot ensure that mistakes won’t threaten civilians’ safety.
The fixes on the table
Vals wants AI developers to strengthen safeguards and train models to resist misuse involving conventional weapons. The concern extends beyond its findings. Anthropic’s research also found that AI models could assist with tactical intelligence work. Anthropic says it has introduced classifiers to detect and block misuse.
Governments have a role as well. SIPRI recommends setting clear requirements for military AI suppliers. It also urges governments to test suppliers’ claims independently and establish accountability before awarding contracts.
A growing market and a Pentagon feud
Military AI is also attracting more investment. Grand View Research estimates the market at $11.7 billion in 2026 and expects it to reach $19.3 billion by 2030. As Cryptopolitan reported, the Pentagon aims to establish AutoWarCom in October 2027.
However, the Pentagon’s partnerships with AI companies have also sparked disputes. A congressional report says Anthropic, Google, OpenAI, and xAI received contracts in 2025 worth up to $200 million each. These amounts are spending limits, not guaranteed payments.
Anthropic opposed using its technology for mass domestic surveillance and fully autonomous weapons. The report also notes that the Pentagon is not publicly known to use any frontier AI model in an autonomous weapon.
Geneva splits over binding rules
Countries are also divided over how to regulate autonomous weapons. At UN talks in Geneva, most of the 128 states party to the Convention on Certain Conventional Weapons support negotiations on binding rules, Reuters reported.
The US, Russia and Israel oppose that approach. They argue that existing humanitarian law already covers autonomous weapons. Russia also delayed the opening discussions over whether observers could speak.
Washington proposed five years of negotiations toward a non-binding agreement. The meeting’s chair suggested a two-year process to develop a binding protocol. Meanwhile, UN Secretary-General António Guterres urged governments to agree on clear prohibitions and restrictions.
The smartest crypto minds already read our newsletter. Want in? Join them.
Article
Microsoft launches six MAI models to challenge rivals on AI costMicrosoft showcased six MAI models at its Ignite conference on November 17. Some were newly launched, while others became more widely available. Microsoft says its models can nearly match more expensive rivals on the tasks they’re built for, while costing far less. Specialized models aimed at the cost of a task Microsoft’s six MAI specialized models include Code 2, Code 2 Flash, Cyber 2 Flash, Image 2.7, Transcribe 2, and Realtime 1. Central to Microsoft’s claims of lower cost at almost the same or better performance is its MAI Code 2 model. The company tested the model using four coding benchmarks: DeepSWE 1.1, Terminal-Bench 2.1, SWE Atlas, and FrontierCode 1.1. In results published on November 17, Microsoft says Code 2 delivered 96% of Claude Opus 5.5’s performance while costing just 5% as much per task. The company also compared Code 2 with GPT-6.1 Sol. This time, it reported 98% of the performance at 40% of the cost. As stated on GitHub’s pricing page, the MAI Code 2 is priced at $0.50 for every million input tokens and $2.50 for every million output tokens. The MAI Code 2 Flash version is available at a lower price of $0.05 for input tokens and $0.25 for output tokens. The Claude Opus 5.5 charges $4 for input and $20 for output, while the GPT-6.1 Sol charges $2 and $10, respectively. However, the prices of Microsoft’s models are only promotional in nature. The costs of the MAI Code 2 and the MAI Code 2 Flash will double after May 17, 2027. Microsoft also reported its model performing better in terms of cybersecurity. Its Cyber 2 Flash achieved a 92.3% cybersecurity score on the CyberGym-E2E benchmark compared to GPT-6 Astra, which got a score of 91.5%. The input/output rates for Microsoft Cyber 2 Flash is $0.10/$0.50 per every million tokens. However, the claims of Microsoft about its models’ performance should be taken with caution. The company chose the benchmarks used to compare its models with rivals. Also, independent testing has yet to confirm the results. Microsoft Launches Six MAI Models to Challenge OpenAI and Anthropic Suleyman frames the tradeoff Mustafa Suleyman, Microsoft’s AI chief, explained the tradeoff in an Axios interview on November 17th. He elaborated that models developed for certain tasks have greater capability and knowledge using fewer parameters. However, there is still a limit to this specialization. A model that performs well at basic coding can have problems in advanced science problems. To address the concern, Suleyman said his team is trying to build “true superintelligence”—an AI system that is capable of performing any intellectual task better than a human being. However, he did not say when such a system might be available. Copilot, Azure weights and platform control Code 2 is already available in GitHub Copilot. That could reduce purchases of outside models. Enterprise customers can also license MAI weights to customize models. However, Axios reported that those customized models must run on Azure, Microsoft’s cloud computing platform. Lower model costs could therefore come with greater cloud dependence. A continuation of Nadella’s supplier worry Microsoft had already introduced seven in-house models on June 2. Its latest releases also address a concern raised by CEO Satya Nadella earlier this year. According to a report by Cryptopolitan on July 17, Nadella raised doubts about the idea of depending on limited leading AI companies for token capacity. By creating its own models, Microsoft is able to secure ample bargaining power while negotiating with such companies. Falling prices, uneven access Microsoft is increasing its portfolio of AI offerings as prices across the industry continue to decrease. According to a study by Epoch AI published on September 22, it has become cheaper to achieve similar AI results since 2023 by as much as 47% quarterly. This translates to approximately 13 times lower price every year. Despite falling costs, established AI providers still dominate business spending. Ramp’s October 2026 index shows Anthropic accounted for 49.3% of token spending among US businesses using its platform. OpenAI followed at 46.1% for the week ending November 1. The figures do not represent the global market. AI infrastructure is also heavily concentrated in a few countries and companies. Stanford’s 2026 AI Index counted 5,427 data centers in the United States in 2025. That’s more than ten times the number in any other country. Stanford also found that Taiwan’s TSMC manufactures almost every leading AI chip. This concentration raises concerns about competition. A July 10 OECD report warned that control over chips and cloud infrastructure could make it harder for other companies to compete in the AI market. Microsoft’s new MAI models raise a similar concern. Lower prices could make its AI models more affordable for businesses worldwide to adopt. But the Azure requirement could also make it harder for customers to switch cloud providers, even if they save money on AI models. The smartest crypto minds already read our newsletter. Want in? Join them.

Microsoft launches six MAI models to challenge rivals on AI cost

Microsoft showcased six MAI models at its Ignite conference on November 17. Some were newly launched, while others became more widely available. Microsoft says its models can nearly match more expensive rivals on the tasks they’re built for, while costing far less.
Specialized models aimed at the cost of a task
Microsoft’s six MAI specialized models include Code 2, Code 2 Flash, Cyber 2 Flash, Image 2.7, Transcribe 2, and Realtime 1.
Central to Microsoft’s claims of lower cost at almost the same or better performance is its MAI Code 2 model. The company tested the model using four coding benchmarks: DeepSWE 1.1, Terminal-Bench 2.1, SWE Atlas, and FrontierCode 1.1. In results published on November 17, Microsoft says Code 2 delivered 96% of Claude Opus 5.5’s performance while costing just 5% as much per task. The company also compared Code 2 with GPT-6.1 Sol. This time, it reported 98% of the performance at 40% of the cost.
As stated on GitHub’s pricing page, the MAI Code 2 is priced at $0.50 for every million input tokens and $2.50 for every million output tokens. The MAI Code 2 Flash version is available at a lower price of $0.05 for input tokens and $0.25 for output tokens. The Claude Opus 5.5 charges $4 for input and $20 for output, while the GPT-6.1 Sol charges $2 and $10, respectively. However, the prices of Microsoft’s models are only promotional in nature. The costs of the MAI Code 2 and the MAI Code 2 Flash will double after May 17, 2027.
Microsoft also reported its model performing better in terms of cybersecurity. Its Cyber 2 Flash achieved a 92.3% cybersecurity score on the CyberGym-E2E benchmark compared to GPT-6 Astra, which got a score of 91.5%. The input/output rates for Microsoft Cyber 2 Flash is $0.10/$0.50 per every million tokens.
However, the claims of Microsoft about its models’ performance should be taken with caution. The company chose the benchmarks used to compare its models with rivals. Also, independent testing has yet to confirm the results.
Microsoft Launches Six MAI Models to Challenge OpenAI and Anthropic
Suleyman frames the tradeoff
Mustafa Suleyman, Microsoft’s AI chief, explained the tradeoff in an Axios interview on November 17th. He elaborated that models developed for certain tasks have greater capability and knowledge using fewer parameters. However, there is still a limit to this specialization. A model that performs well at basic coding can have problems in advanced science problems.
To address the concern, Suleyman said his team is trying to build “true superintelligence”—an AI system that is capable of performing any intellectual task better than a human being. However, he did not say when such a system might be available.
Copilot, Azure weights and platform control
Code 2 is already available in GitHub Copilot. That could reduce purchases of outside models. Enterprise customers can also license MAI weights to customize models.
However, Axios reported that those customized models must run on Azure, Microsoft’s cloud computing platform. Lower model costs could therefore come with greater cloud dependence.
A continuation of Nadella’s supplier worry
Microsoft had already introduced seven in-house models on June 2. Its latest releases also address a concern raised by CEO Satya Nadella earlier this year.
According to a report by Cryptopolitan on July 17, Nadella raised doubts about the idea of depending on limited leading AI companies for token capacity. By creating its own models, Microsoft is able to secure ample bargaining power while negotiating with such companies.
Falling prices, uneven access
Microsoft is increasing its portfolio of AI offerings as prices across the industry continue to decrease. According to a study by Epoch AI published on September 22, it has become cheaper to achieve similar AI results since 2023 by as much as 47% quarterly. This translates to approximately 13 times lower price every year.
Despite falling costs, established AI providers still dominate business spending. Ramp’s October 2026 index shows Anthropic accounted for 49.3% of token spending among US businesses using its platform. OpenAI followed at 46.1% for the week ending November 1. The figures do not represent the global market.
AI infrastructure is also heavily concentrated in a few countries and companies. Stanford’s 2026 AI Index counted 5,427 data centers in the United States in 2025. That’s more than ten times the number in any other country. Stanford also found that Taiwan’s TSMC manufactures almost every leading AI chip.
This concentration raises concerns about competition. A July 10 OECD report warned that control over chips and cloud infrastructure could make it harder for other companies to compete in the AI market.
Microsoft’s new MAI models raise a similar concern. Lower prices could make its AI models more affordable for businesses worldwide to adopt. But the Azure requirement could also make it harder for customers to switch cloud providers, even if they save money on AI models.
The smartest crypto minds already read our newsletter. Want in? Join them.
Article
Coinbase brings Morpho crypto loans to New York in major onchain expansionCoinbase has made its crypto-focused lending service available to residents of New York on Tuesday. New York becomes the last US state in Coinbase’s expansion plans. Its residents can now borrow USDC against Bitcoin, Ether, or Solana using Morpho on Base. What New Yorkers can now borrow Residents of New York can borrow money without selling their crypto assets. Coinbase Borrow has established caps of $5 million for Bitcoin, $1 million for Ethereum, and $100,000 for Solana. These are maximum limits, not necessarily amounts that borrowers will be able to borrow. Eligibility and available collateral are key factors. According to Coinbase, New York’s Department of Financial Services (NYDFS) has approved the offering. Executive Jacob Frantz did not provide any information about application or approval dates or specifics about the approved offering. The NYDFS directory shows that Coinbase received its licenses in 2017 but does not reflect any approval of the new offering. In the meantime, the borrowing page of Coinbase’s website still states that New Yorkers cannot avail of the offering. At the moment, New Yorkers cannot borrow using fixed-rate Morpho Midnight loans, tokenized gold collateral, or the separate USDC yield offered by Coinbase. A regulated front door to onchain credit Customers transact through Coinbase, and Morpho uses smart contracts on Base, which is the Ethereum layer-2 network, to handle loan transactions. There is no need for them to use a separate DeFi interface. Thus, simplifying things for customers. This model has expanded rapidly. Morpho claimed that $3.4 billion was lent through Coinbase Borrow since January of 2025. Currently, there are $1.7 billion outstanding loans, and that amount is secured by at least $3.2 billion worth of collateral. In comparison, data compiled by DeFiLlama indicates that the total outstanding loans currently circulating in Morpho networks amounts to $5.47 billion, as of November 17. Coinbase Morpho Crypto Loans Reach New York: Key Figures How Base lending rates have held up Coinbase’s expansion could bring more borrowers to Base. But whether interest rates remain stable depends on how much money lenders make available. According to Steakhouse Financial, borrowing on Morpho’s cbBTC/USDC market increased from $1.2 billion in June to about $1.6 billion in November. The rates have remained around 4.5%-5%. That balance was tested during the market turmoil surrounding the late-October Binance hack. Spark withdrew roughly $250 million, leaving more than 90% of available funds borrowed. Interest rates jumped to 7%–9.5%. Grove then deposited around $130 million. Within two days, rates had fallen back toward 5.5%. For Coinbase borrowers, the episode shows how quickly lending costs can change when large investors withdraw or add funds. A shrinking market and real risks Borrowers still face fluctuating interest and liquidation risks. Coinbase’s terms warn that an 86% loan-to-value ratio triggers automatic collateral liquidation and a penalty. Cryptopolitan has noted the concentration of Base lending among a few protocols, along with smart-contract, oracle and curator risks. In the second quarter, Galaxy Research estimated crypto-collateralized lending shrank 16.78% to $56.16 billion. Its September outlook pointed to a partial recovery. Federal Reserve Governor Christopher Waller also addressed tokenized markets and payments on Tuesday. He did not mention Coinbase, so the speech does not confirm its New York approval. Coinbase lists availability only in the US and UK; wider expansion remains uncertain. The smartest crypto minds already read our newsletter. Want in? Join them.

Coinbase brings Morpho crypto loans to New York in major onchain expansion

Coinbase has made its crypto-focused lending service available to residents of New York on Tuesday. New York becomes the last US state in Coinbase’s expansion plans. Its residents can now borrow USDC against Bitcoin, Ether, or Solana using Morpho on Base.
What New Yorkers can now borrow
Residents of New York can borrow money without selling their crypto assets. Coinbase Borrow has established caps of $5 million for Bitcoin, $1 million for Ethereum, and $100,000 for Solana. These are maximum limits, not necessarily amounts that borrowers will be able to borrow. Eligibility and available collateral are key factors.
According to Coinbase, New York’s Department of Financial Services (NYDFS) has approved the offering. Executive Jacob Frantz did not provide any information about application or approval dates or specifics about the approved offering.
The NYDFS directory shows that Coinbase received its licenses in 2017 but does not reflect any approval of the new offering. In the meantime, the borrowing page of Coinbase’s website still states that New Yorkers cannot avail of the offering.
At the moment, New Yorkers cannot borrow using fixed-rate Morpho Midnight loans, tokenized gold collateral, or the separate USDC yield offered by Coinbase.
A regulated front door to onchain credit
Customers transact through Coinbase, and Morpho uses smart contracts on Base, which is the Ethereum layer-2 network, to handle loan transactions. There is no need for them to use a separate DeFi interface. Thus, simplifying things for customers.
This model has expanded rapidly. Morpho claimed that $3.4 billion was lent through Coinbase Borrow since January of 2025. Currently, there are $1.7 billion outstanding loans, and that amount is secured by at least $3.2 billion worth of collateral. In comparison, data compiled by DeFiLlama indicates that the total outstanding loans currently circulating in Morpho networks amounts to $5.47 billion, as of November 17.
Coinbase Morpho Crypto Loans Reach New York: Key Figures
How Base lending rates have held up
Coinbase’s expansion could bring more borrowers to Base. But whether interest rates remain stable depends on how much money lenders make available. According to Steakhouse Financial, borrowing on Morpho’s cbBTC/USDC market increased from $1.2 billion in June to about $1.6 billion in November. The rates have remained around 4.5%-5%.
That balance was tested during the market turmoil surrounding the late-October Binance hack. Spark withdrew roughly $250 million, leaving more than 90% of available funds borrowed. Interest rates jumped to 7%–9.5%. Grove then deposited around $130 million. Within two days, rates had fallen back toward 5.5%.
For Coinbase borrowers, the episode shows how quickly lending costs can change when large investors withdraw or add funds.
A shrinking market and real risks
Borrowers still face fluctuating interest and liquidation risks. Coinbase’s terms warn that an 86% loan-to-value ratio triggers automatic collateral liquidation and a penalty. Cryptopolitan has noted the concentration of Base lending among a few protocols, along with smart-contract, oracle and curator risks.
In the second quarter, Galaxy Research estimated crypto-collateralized lending shrank 16.78% to $56.16 billion. Its September outlook pointed to a partial recovery.
Federal Reserve Governor Christopher Waller also addressed tokenized markets and payments on Tuesday. He did not mention Coinbase, so the speech does not confirm its New York approval. Coinbase lists availability only in the US and UK; wider expansion remains uncertain.
The smartest crypto minds already read our newsletter. Want in? Join them.
Coinstancy pits USDC agent wallets against Coinbase and CircleCoinstancy has launched Agentic Bank, a product that gives AI agents their own USDC wallets. It is governed by budgets and approved vendor lists. It also works within the approval limits the owner sets. The product was launched on November 18, and with its release, Coinstancy has entered a market that already has players such as Coinbase, Circle, Binance, and Anchorage. The crypto savings company said in its statement that Agentic Bank bundles dedicated agent wallets and spending policies, as well as hosted execution, all in one environment. Businesses and developers whose software already shops for paid data and APIs mid-task are the target for the product. Where does Coinstancy’s Agentic Bank wallet sit? Each agent has its own USDC wallet on Base, Coinbase’s layer-2 network. However, that wallet is not standing alone, as it sits inside the owner’s existing Coinstancy account. It has its own personal Model Context Protocol (MCP) that lets the AI read the owner’s portfolio, transactions, rates and investments. However, human confirmation is needed before any money moves. The company distinguished the difference between the assistant layer and the delegated agent wallet. It stated that the wallet has a record of its own transactions, as well as daily and monthly budgets. Workflows that outlast the chat window According to Coinstancy, the agent can carry out scheduled or event-triggered tasks without the need for the team to keep a chat open through a module called Cloud Run. An owner can observe how a workflow is through a simulation mode. This enables them to understand how the funds will move before the agent gets to work. Agent access is revocable, and users can also pause paid calls. Who is already playing in the market? Some notable and established companies already have similar products in the market, with some of the launches in the first quarter of 2026. One of them is Coinbase, which launched Agentic Wallets in February. The product comes with an agent that has a standalone wallet that is designed to hold USDC and trade on Base. The session and per-transaction caps and keys are held in Coinbase enclaves. Circle, the USDC issuer, released its own product called Agent Stack on May 11. It is a multichain set of wallets, a CLI, and a 600-service marketplace spanning nine mainnets. Binance entered the market in August with Agent OS in August, wiring tools like ChatGPT and Claude into its markets through a dedicated sub-account walled off from user funds. There is another product that has a nearly similar name to Coinstancy Agentic Bank, and that is Anchorage’s Agentic Banking. However, the product is still operating at its beta stage. It is built on a federally chartered bank to route agent payments across stablecoins, ACH, and wires. What is the current challenge with these agents? So far, recent occurrences have shown that the controls for these agents are not airtight. A recent study by Cornell Tech, Yale, and the Initiative for CryptoCurrencies and Contracts found that nine of 10 frontier AI models given access to a user’s inbox raised their own spending cap on a Circle agent wallet at least once, reading the one-time passcode Circle emails to approve such changes. Only one model, claude-opus-5.5, never exceeded its limit across 20 runs. In another test, five of the 10 models paid an attacker who posed as the owner by email. Some of the builders, such as Circle, have already started putting measures in place to mitigate these issues. Circle in particular now warns owners not to give the agent unrestricted access to their inbox. Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free.

Coinstancy pits USDC agent wallets against Coinbase and Circle

Coinstancy has launched Agentic Bank, a product that gives AI agents their own USDC wallets. It is governed by budgets and approved vendor lists. It also works within the approval limits the owner sets.
The product was launched on November 18, and with its release, Coinstancy has entered a market that already has players such as Coinbase, Circle, Binance, and Anchorage.
The crypto savings company said in its statement that Agentic Bank bundles dedicated agent wallets and spending policies, as well as hosted execution, all in one environment.
Businesses and developers whose software already shops for paid data and APIs mid-task are the target for the product.
Where does Coinstancy’s Agentic Bank wallet sit?
Each agent has its own USDC wallet on Base, Coinbase’s layer-2 network. However, that wallet is not standing alone, as it sits inside the owner’s existing Coinstancy account.
It has its own personal Model Context Protocol (MCP) that lets the AI read the owner’s portfolio, transactions, rates and investments.
However, human confirmation is needed before any money moves. The company distinguished the difference between the assistant layer and the delegated agent wallet. It stated that the wallet has a record of its own transactions, as well as daily and monthly budgets.
Workflows that outlast the chat window
According to Coinstancy, the agent can carry out scheduled or event-triggered tasks without the need for the team to keep a chat open through a module called Cloud Run.
An owner can observe how a workflow is through a simulation mode. This enables them to understand how the funds will move before the agent gets to work.
Agent access is revocable, and users can also pause paid calls.
Who is already playing in the market?
Some notable and established companies already have similar products in the market, with some of the launches in the first quarter of 2026.
One of them is Coinbase, which launched Agentic Wallets in February. The product comes with an agent that has a standalone wallet that is designed to hold USDC and trade on Base. The session and per-transaction caps and keys are held in Coinbase enclaves.
Circle, the USDC issuer, released its own product called Agent Stack on May 11. It is a multichain set of wallets, a CLI, and a 600-service marketplace spanning nine mainnets.
Binance entered the market in August with Agent OS in August, wiring tools like ChatGPT and Claude into its markets through a dedicated sub-account walled off from user funds.
There is another product that has a nearly similar name to Coinstancy Agentic Bank, and that is Anchorage’s Agentic Banking. However, the product is still operating at its beta stage. It is built on a federally chartered bank to route agent payments across stablecoins, ACH, and wires.
What is the current challenge with these agents?
So far, recent occurrences have shown that the controls for these agents are not airtight. A recent study by Cornell Tech, Yale, and the Initiative for CryptoCurrencies and Contracts found that nine of 10 frontier AI models given access to a user’s inbox raised their own spending cap on a Circle agent wallet at least once, reading the one-time passcode Circle emails to approve such changes.
Only one model, claude-opus-5.5, never exceeded its limit across 20 runs. In another test, five of the 10 models paid an attacker who posed as the owner by email.
Some of the builders, such as Circle, have already started putting measures in place to mitigate these issues. Circle in particular now warns owners not to give the agent unrestricted access to their inbox.
Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free.
Verified
CoinGecko could become Solana’s next big trading gatewayThe Solana Foundation has acquired a majority stake in crypto data platform CoinGecko, allowing users to trade on Solana on one of the industry’s most popular platforms, chief product officer Vibhu Norby said Tuesday. CoinGecko confirmed that it received investment from Solana, although no details on the amount were provided. Neither party has provided details on the investment amount or the stake’s price. The news was announced at Solana’s Breakpoint conference in London and included plans for the foundation to develop “new on-platform trading and DeFi experiences with Solana.” Norby wrote on X. The investment will be used to improve data infrastructure, increase coverage of tokenized real-world assets and derivatives, and on-chain trading and intelligence tools. Trading on the data platform The main idea behind this partnership is that CoinGecko users will be able to trade in-app. Both parties are currently looking to introduce Solana trading on the CoinGecko website, GeckoTerminal platform and mobile app. While this is still being explored, it means users can research tokens, track prices and trade in one place. Norby explained that the roadmap includes DeFi tools and increased coverage of stocks and tokenized real-world assets. CoinGecko explained that they want to increase coverage of derivatives and blockchain-based assets. Solana eyes CoinGecko’s massive crypto audience CoinGecko has assured users that the data will still remain open. They explained that methods will be kept transparent and available to the public. The website, app, and application programming interface (API) will continue to operate the same. This is a large reach, as the platform states they have over 10 million monthly users and 200 million+ monthly page views, according to their About page, since they were founded in 2014. They reportedly have 18 million registered users. This means new Solana products in the future will be available to a large audience, as they have around 180,000 applications and businesses using their data feeds. Similar to Binance and CoinMarketCap This partnership is similar to Binance buying a competitor aggregator platform, CoinMarketCap, back in 2020. This partnership comes after an overall increase in institutional support for Solana. At Breakpoint, Solana Foundation President Lily Liu explained how the TRUMP memecoin launch in January 2025 and the liquidation wave in October 2025 were real-time stress tests for the network. She reportedly stated that validators and liquidity pools continued operating throughout the weekend during the traditional market close. During the selloff, where around $20 billion was lost in positions in 2 hours, Solana saw a peak of 60,000 TPS with zero downtime. The Depository Trust & Clearing Corporation, Schroders, and Formation Securities are all progressing with tokenization strategies related to the network. Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free.

CoinGecko could become Solana’s next big trading gateway

The Solana Foundation has acquired a majority stake in crypto data platform CoinGecko, allowing users to trade on Solana on one of the industry’s most popular platforms, chief product officer Vibhu Norby said Tuesday.
CoinGecko confirmed that it received investment from Solana, although no details on the amount were provided. Neither party has provided details on the investment amount or the stake’s price. The news was announced at Solana’s Breakpoint conference in London and included plans for the foundation to develop “new on-platform trading and DeFi experiences with Solana.” Norby wrote on X.
The investment will be used to improve data infrastructure, increase coverage of tokenized real-world assets and derivatives, and on-chain trading and intelligence tools.
Trading on the data platform
The main idea behind this partnership is that CoinGecko users will be able to trade in-app. Both parties are currently looking to introduce Solana trading on the CoinGecko website, GeckoTerminal platform and mobile app. While this is still being explored, it means users can research tokens, track prices and trade in one place.
Norby explained that the roadmap includes DeFi tools and increased coverage of stocks and tokenized real-world assets. CoinGecko explained that they want to increase coverage of derivatives and blockchain-based assets.
Solana eyes CoinGecko’s massive crypto audience
CoinGecko has assured users that the data will still remain open. They explained that methods will be kept transparent and available to the public. The website, app, and application programming interface (API) will continue to operate the same.
This is a large reach, as the platform states they have over 10 million monthly users and 200 million+ monthly page views, according to their About page, since they were founded in 2014. They reportedly have 18 million registered users. This means new Solana products in the future will be available to a large audience, as they have around 180,000 applications and businesses using their data feeds.
Similar to Binance and CoinMarketCap
This partnership is similar to Binance buying a competitor aggregator platform, CoinMarketCap, back in 2020.
This partnership comes after an overall increase in institutional support for Solana. At Breakpoint, Solana Foundation President Lily Liu explained how the TRUMP memecoin launch in January 2025 and the liquidation wave in October 2025 were real-time stress tests for the network.
She reportedly stated that validators and liquidity pools continued operating throughout the weekend during the traditional market close. During the selloff, where around $20 billion was lost in positions in 2 hours, Solana saw a peak of 60,000 TPS with zero downtime. The Depository Trust & Clearing Corporation, Schroders, and Formation Securities are all progressing with tokenization strategies related to the network.
Don’t just read crypto news. Understand it. Subscribe to our newsletter. It's free.
Perplexity says it hit $1B run rate on a fifth of Anthropic's burnPerplexity CEO Aravind Srinivas says the AI search startup has hit a $1 billion annualized revenue run rate while spending about five times less cash than Anthropic. Srinivas made the claim just days before Anthropic is due to start trading on the Nasdaq with the ANT ticker.  What did Perplexity’s CEO claim? Aravind Srinivas, Perplexity’s CEO, spoke on Monday at Semafor’s Silicon Valley & The World event. He told editor-in-chief Ben Smith that his company reached an annualized revenue run rate of $1 billion that same day.  “We spent 5x lower than the company that’s going to IPO this week to get to that number,” he said, according to remarks Semafor posted on X. The only company going public this week is Anthropic, which opens trading on Thursday. Semafor noted that Perplexity does not train its own flagship AI model. It mostly sends user questions to outside systems, including OpenAI’s ChatGPT and Anthropic’s Claude. That system means that some of Perplexity’s spending ends up as revenue for Anthropic.  Srinivas has said that the setup Perplexity uses is a strength and stated that users care less about what model gave them the answer and only “really care if the work actually got done.” Notably, Perplexity’s annualized revenue was below $250 million at the start of 2026. Cryptopolitan reported that it passed $750 million by August due to its Perplexity Computer agent gaining users. The company recorded a roughly 4x rise this year.  Srinivas said Perplexity spent much less than Anthropic to reach $1 billion in annualized revenue due to its capital-efficient business model. The company was last valued at about $20 billion in a funding round in September 2025. As of mid-2026, it had raised about $900 million in total, much less than its bigger rivals. How big is Anthropic’s stock listing? Anthropic and some of its shareholders plan to sell 825 million shares at $70 to $78 each, raising as much as $64.4 billion. The stock is due to trade on the Nasdaq under the ticker ANT. At the top of the price range, the prospectus values the company at about $1.4 trillion when all possible shares are counted. Without options and restricted shares, the value is closer to $1 trillion. The filing also shows a run-rate revenue of about $81 billion as of October 31, up $10 billion in one month. It lists nearly $900 billion in computing commitments.  Beyond comparing AI spending, Srinivas also praised former Nvidia (NASDAQ: NVDA) CEO Jensen Huang on stage. He called Huang “one of the greatest founders in American history” and said “America needs to have more Jensens.” Huang reportedly once advised him to run a company as if it might fail within a year.  Despite the praise for Huang, when asked if Nvidia’s board should reinstate him, Srinivas dodged the question. The smartest crypto minds already read our newsletter. Want in? Join them.

Perplexity says it hit $1B run rate on a fifth of Anthropic's burn

Perplexity CEO Aravind Srinivas says the AI search startup has hit a $1 billion annualized revenue run rate while spending about five times less cash than Anthropic.
Srinivas made the claim just days before Anthropic is due to start trading on the Nasdaq with the ANT ticker.
What did Perplexity’s CEO claim?
Aravind Srinivas, Perplexity’s CEO, spoke on Monday at Semafor’s Silicon Valley & The World event. He told editor-in-chief Ben Smith that his company reached an annualized revenue run rate of $1 billion that same day.
“We spent 5x lower than the company that’s going to IPO this week to get to that number,” he said, according to remarks Semafor posted on X. The only company going public this week is Anthropic, which opens trading on Thursday.
Semafor noted that Perplexity does not train its own flagship AI model. It mostly sends user questions to outside systems, including OpenAI’s ChatGPT and Anthropic’s Claude. That system means that some of Perplexity’s spending ends up as revenue for Anthropic.
Srinivas has said that the setup Perplexity uses is a strength and stated that users care less about what model gave them the answer and only “really care if the work actually got done.”
Notably, Perplexity’s annualized revenue was below $250 million at the start of 2026. Cryptopolitan reported that it passed $750 million by August due to its Perplexity Computer agent gaining users. The company recorded a roughly 4x rise this year.
Srinivas said Perplexity spent much less than Anthropic to reach $1 billion in annualized revenue due to its capital-efficient business model. The company was last valued at about $20 billion in a funding round in September 2025. As of mid-2026, it had raised about $900 million in total, much less than its bigger rivals.
How big is Anthropic’s stock listing?
Anthropic and some of its shareholders plan to sell 825 million shares at $70 to $78 each, raising as much as $64.4 billion. The stock is due to trade on the Nasdaq under the ticker ANT. At the top of the price range, the prospectus values the company at about $1.4 trillion when all possible shares are counted. Without options and restricted shares, the value is closer to $1 trillion.
The filing also shows a run-rate revenue of about $81 billion as of October 31, up $10 billion in one month. It lists nearly $900 billion in computing commitments.
Beyond comparing AI spending, Srinivas also praised former Nvidia (NASDAQ: NVDA) CEO Jensen Huang on stage. He called Huang “one of the greatest founders in American history” and said “America needs to have more Jensens.” Huang reportedly once advised him to run a company as if it might fail within a year.
Despite the praise for Huang, when asked if Nvidia’s board should reinstate him, Srinivas dodged the question.
The smartest crypto minds already read our newsletter. Want in? Join them.
Merlin bridge mix-up leaves an innocent user's $3M in BTC frozen in India, ZachXBT saysAn Indian court has ordered crypto exchange Gate to hand about $3 million in bitcoin to a romance scam victim, even though the coins belong to an investor who was not part of the fraud. Gate says it has not moved any funds and will review the case on its own terms. The investor’s account has been frozen for about 16 months. Will Gate hand over $3 million in bitcoin? During the course of investigating a romance scam in India that cost victims about $1.7 million, police found that the stolen money was transferred through a crypto bridge. However, instead of treating it as a shared tool, officers labeled the bridge itself as a scam address. ZachXBT, who traces crypto transactions, has called out that action as a mistake. The bridge in question is run by Merlin Chain. They lock bitcoin on the main Bitcoin network to get a matching token, then swap the token back when they want their bitcoin returned. Research firm Spark says Merlin Chain is the biggest Bitcoin layer-2 network by total value locked, at around $1.7 billion. It also says the bridge has handled about $16 billion in transfers since its launch. Due to the police flagging the bridge as a scam address, any withdrawal made through it looked like laundered money. One of those withdrawals was received into the Gate account of an investor from China who ZachXBT says had no link to the $1.7 million romance scam. Due to the transfer from the flagged bridge, the investor’s account was frozen while containing close to $3 million. Now, the court has ordered the bitcoin to be given to the person who reported the scam, rather than returned to the investor. Gate posted a statement in Chinese on November 15, saying that it is an independent third party and does not decide who owns disputed funds. It also said it will not release assets based on one side’s paperwork. The crypto exchange said it looks at each request case by case, and that no assets have been transferred so far. The case reportedly dates back to 2025. Why India’s crypto freezes are being criticized as excessive This is not the first time Indian police have been accused of being heavy-handed with its freezes. In 2022, police blocked the bank account of crypto exchange Bitbns. Its CEO, Gaurav Dahake, stated that officers often freeze a whole account over a cybercrime complaint worth as little as 10,000 rupees (about $125), instead of freezing only the disputed amount. In January 2026, the Delhi High Court ruled that police need approval from a competent court before freezing bank accounts, and it canceled the freeze orders on accounts held by Malabar Gold and Diamonds. ZachXBT has shared his analysis of how investigators traced the romance scam money into the bridge. But it remains unclear whether the order for Gate to give the money to the complainant will be enforced, challenged on appeal, or thrown out. If you're reading this, you’re already ahead. Stay there with our newsletter.

Merlin bridge mix-up leaves an innocent user's $3M in BTC frozen in India, ZachXBT says

An Indian court has ordered crypto exchange Gate to hand about $3 million in bitcoin to a romance scam victim, even though the coins belong to an investor who was not part of the fraud.
Gate says it has not moved any funds and will review the case on its own terms. The investor’s account has been frozen for about 16 months.
Will Gate hand over $3 million in bitcoin?
During the course of investigating a romance scam in India that cost victims about $1.7 million, police found that the stolen money was transferred through a crypto bridge. However, instead of treating it as a shared tool, officers labeled the bridge itself as a scam address.
ZachXBT, who traces crypto transactions, has called out that action as a mistake. The bridge in question is run by Merlin Chain. They lock bitcoin on the main Bitcoin network to get a matching token, then swap the token back when they want their bitcoin returned.
Research firm Spark says Merlin Chain is the biggest Bitcoin layer-2 network by total value locked, at around $1.7 billion. It also says the bridge has handled about $16 billion in transfers since its launch.
Due to the police flagging the bridge as a scam address, any withdrawal made through it looked like laundered money. One of those withdrawals was received into the Gate account of an investor from China who ZachXBT says had no link to the $1.7 million romance scam.
Due to the transfer from the flagged bridge, the investor’s account was frozen while containing close to $3 million. Now, the court has ordered the bitcoin to be given to the person who reported the scam, rather than returned to the investor.
Gate posted a statement in Chinese on November 15, saying that it is an independent third party and does not decide who owns disputed funds. It also said it will not release assets based on one side’s paperwork.
The crypto exchange said it looks at each request case by case, and that no assets have been transferred so far. The case reportedly dates back to 2025.
Why India’s crypto freezes are being criticized as excessive
This is not the first time Indian police have been accused of being heavy-handed with its freezes. In 2022, police blocked the bank account of crypto exchange Bitbns. Its CEO, Gaurav Dahake, stated that officers often freeze a whole account over a cybercrime complaint worth as little as 10,000 rupees (about $125), instead of freezing only the disputed amount.
In January 2026, the Delhi High Court ruled that police need approval from a competent court before freezing bank accounts, and it canceled the freeze orders on accounts held by Malabar Gold and Diamonds.
ZachXBT has shared his analysis of how investigators traced the romance scam money into the bridge. But it remains unclear whether the order for Gate to give the money to the complainant will be enforced, challenged on appeal, or thrown out.
If you're reading this, you’re already ahead. Stay there with our newsletter.
Ireland’s privacy regulator stops X training AI with data collected by Grok Bot agentsIreland’s Data Protection Commission (DPC) has forced X to stop training its models on data gathered by Grok Bot’s AI agents. It is the second time in about two years that the regulator has cut off a data stream feeding Grok. The first time was in 2024, and the material then was EU users’ own public posts. This time around, it is information that Grok Bot, X’s AI agent, collected while acting on behalf of users across apps and websites. From users’ posts to what agents scoop up about others X suspended the processing after the DPC flagged “significant concerns” in recent weeks, according to Commissioner Dale Sunderland. The so-called third-party data had been mined by Grok Bot since late October and fed into X’s training systems. “We asked them to suspend the processing, and they’ve confirmed to us today that they’ve done that,” Sunderland said in an interview on Monday. Sunderland, who is one of three co-commissioners at the authority, described third-party data as personal information about someone unconnected to the agent’s user. He gave an example to drive home his point: holiday photos a person shares with a friend who then runs an AI agent. It also covers anything that an agent scrapes off the open web.  According to Sunderland, the regulator does not object to AI themselves. Their concern is whether the companies behind these agents stay within the law once they use the data they have gathered to train their models. A regulator that already brought Grok to court once The DPC pushed X to drop EU and EEA users’ public posts from Grok training in 2024. X agreed to suspend processing of posts it had used between May 7 and August 1, 2024, and gave an undertaking to Ireland’s High Court. The commission formally closed the case on September 4, 2024, after X agreed to stop permanently. It was the first time the DPC had gone to court to halt a data practice. However, the watchdog did not rest on its laurels the following year as it opened a separate inquiry into X Internet Unlimited Company over whether training Grok on public posts complied with the GDPR. The DPC can levy fines of up to 4% of a company’s global revenue, and its past penalties against Meta total close to 3 billion euros. Ireland is a member of the EU and its capital, Dublin, hosts the European headquarters of big tech companies, including Apple, Google, Meta, OpenAI and X, among others. Given its position in the bloc, actions that its regulators carry out have far-reaching consequences, as they could set the tone across the bloc. Sunderland said the DPC is also talking to another, unnamed tech firm about plans to train on agent-collected data. “It’s a very live issue at the moment,” he said. Agents that log in as you are already picking fights Ireland is not alone in its concern about the capabilities of agents that sign in on a user’s behalf. Even companies are now wary of the issues. Amazon reportedly blocked Grok Bot from its store, and Amazon’s robots.txt file shuts out agents tied to Grok, ChatGPT, Claude, Copilot, Perplexity, Manus and Google’s crawlers.  Amazon also sued Perplexity, accusing it of quietly reaching into private customer accounts through its Comet browser agent, a claim Perplexity disputes. The DPC is not leaving engagement off the table with X, as it says it will engage with it more on what happened and what comes next. If you're reading this, you’re already ahead. Stay there with our newsletter.

Ireland’s privacy regulator stops X training AI with data collected by Grok Bot agents

Ireland’s Data Protection Commission (DPC) has forced X to stop training its models on data gathered by Grok Bot’s AI agents.
It is the second time in about two years that the regulator has cut off a data stream feeding Grok.
The first time was in 2024, and the material then was EU users’ own public posts. This time around, it is information that Grok Bot, X’s AI agent, collected while acting on behalf of users across apps and websites.
From users’ posts to what agents scoop up about others
X suspended the processing after the DPC flagged “significant concerns” in recent weeks, according to Commissioner Dale Sunderland.
The so-called third-party data had been mined by Grok Bot since late October and fed into X’s training systems.
“We asked them to suspend the processing, and they’ve confirmed to us today that they’ve done that,” Sunderland said in an interview on Monday.
Sunderland, who is one of three co-commissioners at the authority, described third-party data as personal information about someone unconnected to the agent’s user.
He gave an example to drive home his point: holiday photos a person shares with a friend who then runs an AI agent. It also covers anything that an agent scrapes off the open web.
According to Sunderland, the regulator does not object to AI themselves. Their concern is whether the companies behind these agents stay within the law once they use the data they have gathered to train their models.
A regulator that already brought Grok to court once
The DPC pushed X to drop EU and EEA users’ public posts from Grok training in 2024. X agreed to suspend processing of posts it had used between May 7 and August 1, 2024, and gave an undertaking to Ireland’s High Court.
The commission formally closed the case on September 4, 2024, after X agreed to stop permanently. It was the first time the DPC had gone to court to halt a data practice.
However, the watchdog did not rest on its laurels the following year as it opened a separate inquiry into X Internet Unlimited Company over whether training Grok on public posts complied with the GDPR.
The DPC can levy fines of up to 4% of a company’s global revenue, and its past penalties against Meta total close to 3 billion euros.
Ireland is a member of the EU and its capital, Dublin, hosts the European headquarters of big tech companies, including Apple, Google, Meta, OpenAI and X, among others.
Given its position in the bloc, actions that its regulators carry out have far-reaching consequences, as they could set the tone across the bloc.
Sunderland said the DPC is also talking to another, unnamed tech firm about plans to train on agent-collected data. “It’s a very live issue at the moment,” he said.
Agents that log in as you are already picking fights
Ireland is not alone in its concern about the capabilities of agents that sign in on a user’s behalf. Even companies are now wary of the issues. Amazon reportedly blocked Grok Bot from its store, and Amazon’s robots.txt file shuts out agents tied to Grok, ChatGPT, Claude, Copilot, Perplexity, Manus and Google’s crawlers.
Amazon also sued Perplexity, accusing it of quietly reaching into private customer accounts through its Comet browser agent, a claim Perplexity disputes.
The DPC is not leaving engagement off the table with X, as it says it will engage with it more on what happened and what comes next.
If you're reading this, you’re already ahead. Stay there with our newsletter.
Sony beats UK lawsuit over Playstation store pricesSony has won a UK lawsuit over PlayStation Store prices. The case sought about £2 billion ($2.64 billion) in compensation for an estimated 12.2 million gamers. The lawsuit accused the company of stopping other companies from selling digital PlayStation games so it could charge higher prices. However, London’s Competition Appeal Tribunal dismissed the case in a ruling published on Tuesday. This means PlayStation users will not receive any compensation, and the court did not find that Sony broke competition laws. Why the tribunal threw out the claim The tribunal rejected the entire case against Sony. According to Yahoo Finance, the judges said the way the people who filed the lawsuit described the market was “unsafe” and could not be used to decide whether Sony had broken the law. The judges were also not convinced that Sony had too much control over the market, kept competitors out or charged customers unfair prices. The tribunal said people generally had enough information to make their own decisions before buying PlayStation consoles. It said buyers “would be likely to make informed choices when purchasing consoles.” It also said people who spent a lot on games “had both access to sufficient information and an incentive to consider that information” before buying. The ruling came after a 30-day hearing that began in March, according to Reuters. Gamers accused Sony of stifling competition Consumer advocate Alex Neill filed the lawsuit on behalf of about 12.2 million PlayStation users in the UK. According to Yahoo Finance, the lawsuit accused Sony of blocking “all actual and potential competition” in digital PlayStation games. It also claimed the company charged “excessive and unfair” fees for games bought through the PlayStation Store. Robert Palmer KC, the lawyer for the gamers, said Sony did not allow game publishers to sell digital PlayStation games through other stores without its approval. He also claimed PlayStation consoles were built to stop users from downloading games from other stores. According to Yahoo Finance, Palmer said this allowed Sony to set the prices of all digital PlayStation games without facing competition from other stores. The gamers argued that this allowed the company to charge higher prices and make more money than it could if other stores were allowed to compete, Reuters reported. Sony’s defence of a closed store Sony said it had done nothing wrong by selling digital PlayStation games only through its own store. It argued that allowing other stores to sell games on PlayStation could put users’ personal information and security at risk, according to Yahoo Finance. Sony also said it makes very little money from selling PlayStation consoles. It keeps prices low to attract more players and earns back the money through sales on the PlayStation Store. The company said it already competes with Nintendo and Microsoft’s Xbox. It also argued that letting other stores sell PlayStation games would allow them to make money from a system the company had spent money building, Reuters reported. A test for UK collective actions According to Reuters, Sony’s case was the third lawsuit against a major tech company to go to trial in the UK since the start of 2025. New rules have made it easier for large groups of consumers to sue companies together. Other lawsuits are still going on. Yahoo Finance reported that Google faces a £1 billion lawsuit over fees on its Play Store. Apple is also facing several lawsuits worth billions of pounds. Despite Sony’s win, big tech companies continue to face lawsuits over the way they run their app stores and sell digital content. The smartest crypto minds already read our newsletter. Want in? Join them.

Sony beats UK lawsuit over Playstation store prices

Sony has won a UK lawsuit over PlayStation Store prices. The case sought about £2 billion ($2.64 billion) in compensation for an estimated 12.2 million gamers.
The lawsuit accused the company of stopping other companies from selling digital PlayStation games so it could charge higher prices.
However, London’s Competition Appeal Tribunal dismissed the case in a ruling published on Tuesday. This means PlayStation users will not receive any compensation, and the court did not find that Sony broke competition laws.
Why the tribunal threw out the claim
The tribunal rejected the entire case against Sony. According to Yahoo Finance, the judges said the way the people who filed the lawsuit described the market was “unsafe” and could not be used to decide whether Sony had broken the law.
The judges were also not convinced that Sony had too much control over the market, kept competitors out or charged customers unfair prices.
The tribunal said people generally had enough information to make their own decisions before buying PlayStation consoles. It said buyers “would be likely to make informed choices when purchasing consoles.”
It also said people who spent a lot on games “had both access to sufficient information and an incentive to consider that information” before buying. The ruling came after a 30-day hearing that began in March, according to Reuters.
Gamers accused Sony of stifling competition
Consumer advocate Alex Neill filed the lawsuit on behalf of about 12.2 million PlayStation users in the UK.
According to Yahoo Finance, the lawsuit accused Sony of blocking “all actual and potential competition” in digital PlayStation games. It also claimed the company charged “excessive and unfair” fees for games bought through the PlayStation Store.
Robert Palmer KC, the lawyer for the gamers, said Sony did not allow game publishers to sell digital PlayStation games through other stores without its approval.
He also claimed PlayStation consoles were built to stop users from downloading games from other stores. According to Yahoo Finance, Palmer said this allowed Sony to set the prices of all digital PlayStation games without facing competition from other stores.
The gamers argued that this allowed the company to charge higher prices and make more money than it could if other stores were allowed to compete, Reuters reported.
Sony’s defence of a closed store
Sony said it had done nothing wrong by selling digital PlayStation games only through its own store. It argued that allowing other stores to sell games on PlayStation could put users’ personal information and security at risk, according to Yahoo Finance.
Sony also said it makes very little money from selling PlayStation consoles. It keeps prices low to attract more players and earns back the money through sales on the PlayStation Store.
The company said it already competes with Nintendo and Microsoft’s Xbox. It also argued that letting other stores sell PlayStation games would allow them to make money from a system the company had spent money building, Reuters reported.
A test for UK collective actions
According to Reuters, Sony’s case was the third lawsuit against a major tech company to go to trial in the UK since the start of 2025. New rules have made it easier for large groups of consumers to sue companies together.
Other lawsuits are still going on. Yahoo Finance reported that Google faces a £1 billion lawsuit over fees on its Play Store. Apple is also facing several lawsuits worth billions of pounds.
Despite Sony’s win, big tech companies continue to face lawsuits over the way they run their app stores and sell digital content.
The smartest crypto minds already read our newsletter. Want in? Join them.
SONYUS-0.28%
Log in to explore more content
Join global crypto users on Binance Square
⚡️ Get latest and useful information about crypto.
💬 Trusted by the world’s largest crypto exchange.
👍 Discover real insights from verified creators.
Email / Phone number
Sitemap
Cookie Preferences
Platform T&Cs