Coin World Report:
Stability AI has released an improved community license for its Stable Diffusion 3 (SD3) model in an attempt to quell the controversy that erupted after its initial release. The company made this move following the ban imposed by the major community hub, CivitAI, which prohibited all content related to SD3 due to licensing issues.
In a statement released on Friday evening, Stability acknowledged that their latest version, SD3 Medium, did not meet the high expectations of their community. “We have listened to your feedback and made improvements to address your concerns and continue supporting the open-source community,” they said.
According to the new terms, Stability AI allows SD3 to be used free of charge for research, non-commercial, and limited commercial purposes. The license also permits individuals and businesses with annual revenue below $1 million to use the model for free. Those exceeding this threshold must obtain a paid enterprise license.
In an interview with Stability, the company confirmed that it is possible to create customized SD3 models and make improvements upon the underlying SD3. However, it is prohibited to use images generated by SD3 as part of its training dataset to develop new competing models for stable diffusion, using materials from the original model for training purposes.
A spokesperson for Stability AI told Decrypt, “Derivative works include any output derived from Stability AI base models, such as fine-tuned models or other creative outputs. Examples of derivative works include SD3 fine-tuning, LoRA fine-tuning, adapters, etc., which can also be trained using SD3 output images.”
The license also stipulates, “You are the owner of derivative works you create, subject to Stability AI’s ownership of Stability AI materials and any derivative works made by or for Stability AI.” In other words, as long as these boundaries are respected, fine-tuning and profiting from it should not violate the terms and conditions.
The company spokesperson told Decrypt, “To protect our intellectual property, using SD3 outputs as training data to train new underlying AI models is not allowed, and all activities must comply with our acceptable use policy.”
Stability AI claims that the updated license reaffirms the principles of open-source, emphasizing creator freedom, research consistency, and community focus. The company promises to remove the limit on the number of generated media files and assures users below the revenue threshold that they do not need to delete their outputs or fine-tuned models.
To top it off, the company commits to updating its model to address the issues users discovered in the first version, particularly the catastrophic descriptions of human anatomical structures that resulted in monstrous creations when asked to generate a person lying on the floor.
The spokesperson told Decrypt, “We are continuously researching improvements in various technical aspects, and we will release these details with the next model release.”
“One-star” Reviews
Despite Stability AI’s efforts to smooth things over, not everyone in the AI community believes it. Kent Keirsey, CEO of Invoke AI, expressed skepticism about the new license. Keirsey believes that the modified terms fail to address the fundamental issues and introduce new complexities.
In his analysis of the new terms and conditions on YouTube, he stated, “I have to say the new license for SD3 is a bit of an old problem. I know it’s being touted as solving problems, but it doesn’t really solve the problems that I see for people who really care about building a business around this stuff.”
Keirsey pointed out several issues with the license, including Stability AI’s ability to terminate the agreement at will. While this may be normal for many software licenses, it is not part of the most important open-source licenses like GPL or Apache licenses. In fact, the license for the previous SDXL model was irrevocable.
“I would give this one star,” he concluded, “It’s a one-star license, maybe one and a half stars because it’s better than before.”
Edited by Andrew Hayward
