California’s Training Data Transparency Act moves to Senate floor

TCAI co-founder Jai Jaisimha testifying: “We know more about what’s in this pack of gum than we know about AI models.” The Training Data Transparency Act, sponsored by California Assemblymember Jacqui Irwin, right, passed out of the State Senate Judiciary Committee yesterday. It now heads to the Senate floor.

California’s proposed proposed Training Data Transparency Act represents one of the first serious efforts to enact safeguards around the data used to train AI models like OpenAI’s ChatGPT, and it moved one step closer to adoption yesterday.

After hearing testimony from Transparency Coalition Co-founder Jai Jaisimha and Assemblymember Jacqui Irwin (D-Thousand Oaks), the bill’s sponsor, the Senate Judiciary Committee approved the measure and sent it to the Senate floor. A vote by the full Senate is expected within the next few weeks.

AB 2013 requires an AI model developer to post a high-level summary of the datasets used in the development of the AI system. That would include the sources or owners of the datasets, the number of data points in the datasets, whether the datasets were purchased or licensed by the developer, and whether the datasets include personal information.

“Transparency implies openness and accountability,” Jaisimha told members of the Judiciary Committee. “Currently we are seeing neither openness nor accountability from the largest developers of generative AI models, and they are setting the benchmark for everyone else.”

“You will hear that we can’t do this because training data is a ‘trade secret,’” Jaisimha added. “This in our opinion is a false narrative. We as citizens know more what goes into this pack of gum than we do about our AI models. Generative AI models are largely built using the same data. They are different because of all the ingenuity employed in training them. This bill does not require the disclosure of those proprietary methods. And ask yourself, if training data is such a secret, why does the industry insist on loudly announcing every single licensing deal?”

AB 2013 passed out of the Judiciary Committee yesterday on a vote of 10-1. The measure was approved by the full Assembly last month. It next goes before the full Senate, for debate and a vote that has yet to be scheduled. TCAI will be there to support and cover the bill’s progress. A full analysis of AB 2013 is available here.