California Lawmakers Renew Push for Tough AI Training Data Disclosure
California lawmakers advance AB 412 to require AI training data copyright disclosures.
Why it matters: This legislation affects legal tech developers and corporate compliance teams by imposing strict, technically challenging transparency demands on AI training datasets.
- AB 412 introduced Feb 4, 2025, and remains active as of May 28, 2026.
- Requires AI developers to identify copyrighted materials used in training datasets.
- Mandates rights owners can request details on copyrighted data use.
- Industry groups like SIIA oppose AB 412 citing infeasibility and compliance burdens.
California Assembly Bill 412 (AB 412) was introduced by Assembly Member Bauer-Kahan on February 4, 2025 and is under active consideration as of May 2026. The bill would require developers of generative AI models to document copyrighted works included in their training datasets and provide a formal mechanism allowing rights holders to request such information.
This legislative move follows the enactment of California Assembly Bill 2013 (AB 2013), effective January 1, 2026, which mandates AI developers to publicly disclose data sources, copyright status, and personal information inclusion in their training sets. AB 2013 requires this information to be posted on developers' websites before public release and updated after substantial model changes, as analyzed in Crowell & Moring’s legal alert and further explained in Davis Graham & Stubbs LLP’s update.
However, AB 412 raises the bar by demanding AI developers identify any copyrighted works they know were used, which industry groups like the Software & Information Industry Association (SIIA) strongly oppose. SIIA warns that these requirements are technically infeasible, impose significant compliance burdens, risk leaking proprietary information, and could hamper California’s leadership in AI innovation.
Despite enforcement details and penalties for AB 412 not being publicly specified, the renewed legislative push signals California’s intent to assert stringent AI governance standards that may set precedents for legal tech and corporate compliance frameworks nationwide.
By the numbers:
- February 4, 2025 — AB 412 introduction date
- May 28, 2026 — AB 412 status as in progress
- January 1, 2026 — AB 2013 effective date
Yes, but: While AB 412 aims to increase transparency, critics highlight its technical infeasibility and risk to proprietary data, raising questions about practical enforcement and innovation impact.
What's next: As AB 412 progresses through California’s legislative process in 2026, stakeholders are closely watching for amendments, enforcement provisions, and final passage decisions.