I read through the user agreements on the Model Hub line by line, and that’s when I realized that $OPG deliberately concealed a major compliance risk, effectively blocking the commercialization path for the B-side.
The platform fully opens up upload permissions. Anyone can directly upload AI model weights. The system does not automatically verify open-source license agreements, and there is no manual copyright review. Most of the models in the repository are copied over from the open web. Many files don’t even include basic license text like MIT or Apache—living for the long term in the gray area of intellectual property.
The most critical point is how responsibility is allocated. The agreement clearly states in black and white: all copyright disputes are borne solely by the model publisher, and @OpenGradient bears no liability whatsoever, whether direct or joint. The project only builds a storage relay channel and passes all legal and compliance risks to the users.
I also noticed that #opg ’s main customers are heavily regulated enterprises, such as those in finance risk control and healthcare AI. If a company uses an unauthorized model in its business and then gets sued by the copyright holder, all losses must be covered by the company itself; the platform won’t provide any safety net. I think any company with even a moderate scale would never risk integrating such a network.
To make matters worse, the model data is stored in Walrus for permanent storage. Once an infringing file is written on-chain, it becomes extremely difficult to remove completely. Historically non-compliant models will be preserved permanently, and intellectual property risks will only accumulate over time, with legal disputes potentially erupting at any moment.
I reviewed the repository data: among more than four thousand models, there are very few compliance and business-use models developed in-house by the team. Most are unauthorized copies and transfers. They advertise externally that they are a highly trustworthy and auditable enterprise AI infrastructure, yet they allow copyright chaos to go unchecked—doing no risk control throughout.
Verifiable technology is just icing on the cake. When companies do business, the first thing they must avoid is legal risk. The platform only distributes traffic. It’s unwilling to control the origin of the copyrights, and shifts all compliance pressure onto users. With just this pile of copyright “landmines,” it’s hard to attract stable paying developers and enterprise clients. No matter how pretty the decentralized narrative sounds, in the face of real, tangible legal risk, it will lose its appeal. Don’t you think so?
The platform fully opens up upload permissions. Anyone can directly upload AI model weights. The system does not automatically verify open-source license agreements, and there is no manual copyright review. Most of the models in the repository are copied over from the open web. Many files don’t even include basic license text like MIT or Apache—living for the long term in the gray area of intellectual property.
The most critical point is how responsibility is allocated. The agreement clearly states in black and white: all copyright disputes are borne solely by the model publisher, and @OpenGradient bears no liability whatsoever, whether direct or joint. The project only builds a storage relay channel and passes all legal and compliance risks to the users.
I also noticed that #opg ’s main customers are heavily regulated enterprises, such as those in finance risk control and healthcare AI. If a company uses an unauthorized model in its business and then gets sued by the copyright holder, all losses must be covered by the company itself; the platform won’t provide any safety net. I think any company with even a moderate scale would never risk integrating such a network.
To make matters worse, the model data is stored in Walrus for permanent storage. Once an infringing file is written on-chain, it becomes extremely difficult to remove completely. Historically non-compliant models will be preserved permanently, and intellectual property risks will only accumulate over time, with legal disputes potentially erupting at any moment.
I reviewed the repository data: among more than four thousand models, there are very few compliance and business-use models developed in-house by the team. Most are unauthorized copies and transfers. They advertise externally that they are a highly trustworthy and auditable enterprise AI infrastructure, yet they allow copyright chaos to go unchecked—doing no risk control throughout.
Verifiable technology is just icing on the cake. When companies do business, the first thing they must avoid is legal risk. The platform only distributes traffic. It’s unwilling to control the origin of the copyrights, and shifts all compliance pressure onto users. With just this pile of copyright “landmines,” it’s hard to attract stable paying developers and enterprise clients. No matter how pretty the decentralized narrative sounds, in the face of real, tangible legal risk, it will lose its appeal. Don’t you think so?