Legislators in the U.S. have filed a bill requiring creators of foundational AI models to disclose their sources of training data. The proposed AI Foundational Model Transparency Act would direct the Federal Trade Commission to collaborate with the National Institute of Standards and Technology (NIST) to establish rules for reporting training data transparency. Companies producing foundational models will be obligated to reveal the data sources and efforts made to prevent inaccurate or harmful information. This legislation complements the Biden administration’s AI executive order and its focus on AI safety, security, privacy, and civil rights.
The introduction of this bill stems from numerous lawsuits targeting companies that develop and train AI models, accusing them of copyright infringement. One example is the legal action taken by Getty Images against Stability AI, the creators of the text-to-image deep learning model Stable Diffusion. Getty Images alleged that Stability AI unlawfully copied and processed millions of copyrighted images and associated metadata.
In conclusion, the bill proposed by U.S. legislators aims to enhance transparency in the development of foundational AI models by requiring disclosure of training data sources. This move will assist copyright holders in identifying if their data was used and promote responsible AI practices.
