California approves Training Data Transparency Act, now on Gov. Newsom’s desk

In a significant milestone for AI safety and transparency, the California legislature yesterday gave final approval to AB 2013, the Training Data Transparency Act. The Act, authored by Assemblymember Jacqui Irwin, represents one of the first serious efforts to enact safeguards around the data used to train AI models. The Assembly approved the bill on a vote of 75-0, following the Senate’s 38-0 approval on Monday.

AB 2013 has been one of the Transparency Coalition’s top priority bills in the California legislature this year. TCAI founders Jai Jaisimha and Rob Eleveld offered expert testimony at committee hearings and worked closely with the bill’s sponsors, offering suggestions to strengthen and clarify the bill’s intentions.

The approved bill now awaits the signature of Gov. Gavin Newsom.

Key component: High-level summaries of datasets

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. Further details and requirements can be found in the final bill.

The new requirements would take effect on Jan. 1, 2026, giving AI developers nearly 18 months to prepare for the transparency disclosures.

building consumer confidence in ethical AI

“Consumer confidence in AI systems has not grown at the same rapid pace as industry adoption,” Assemblymember Irwin noted earlier this year. “Many consumers have valid questions about how these AI systems and services are created, and if they truly are better than what they seek to replace.”

 “To build consumer confidence,” she added, “we need to start with the foundations, and for AI that is the selection of training data. AB 2013 provides transparency to consumers of AI systems and services by providing important documentation about the data used to train the services and systems they are being offered, including if synthetic data has or is being used to fill gaps in data sources.”

TCAI Chairman Rob Eleveld testified on behalf of the bill earlier this year. The transparency requirements in the Act, he said, “are not heavy-handed requirements, nor will they inhibit small business innovation,” The transparency requirements in Irwin’s bill, he added, will safeguard California’s citizens “while empowering the tech industry to grow and thrive.”

“Now is the time to act,” said Eleveld. “We cannot repeat the missteps and inaction during the nascent stages of social media.”

Previous
Previous

AI Field Notes: California crunch time yields crop of approved AI bills

Next
Next

How AI systems trained on AI-generated data lead to erosion and collapse