AI has faced scrutiny over data usage, particularly as Hollywood grapples with the integration of AI technologies amid copyright disputes. In response, Bria AI—a text-to-image service—has positioned itself as a “responsible” generative AI platform, announcing its latest model trained exclusively on fully licensed data. This approach allows the new model to perform comparably to industry leaders, with notably fewer parameters.

Yair Adato, CEO of Bria, stated, “While the industry races to build ever-larger models using scraped web data, we’ve proven that smaller, ethically-trained models can deliver equivalent performance.” He emphasized that their new model not only respects the rights of creators but also excels within the field.

Bria’s recent release includes an open-sourced version of its model, which can be accessed via Hugging Face, complete with a robust development framework, ControlNets, IP adaptors, and additional auxiliary models. This advancement underscores Bria’s commitment to fostering a sustainable creative ecosystem by directly engaging with artists and content creators to ensure that they benefit from the economic gains driven by AI advancements.

The backdrop for this launch sees industry giants like Getty Images embroiled in legal battles against AI players such as Stability AI, challenging mass copyright theft. The legal proceedings have begun to evolve, with Getty pivoting to focus on trademark infringement and other claims regarding the unauthorized use of its images.

Bria’s initiative is being lauded as a critical step toward establishing ethical frameworks in AI image generation. Vered Horesh, Bria’s Chief Strategy Officer, noted, “Every image generated by Bria represents a vote for a sustainable creative ecosystem.” She emphasized that the company aims to showcase how AI can augment artistic endeavors rather than exploit creators’ contributions.

The new model is available on Bria’s website, marking a significant milestone in the quest for ethical AI development in the image generation landscape.