← Law · AI

AI

California frontier-AI transparency law

A state statute for the largest model developers: publish a safety framework and report serious incidents. Senate Bill 1047 of 2024 was vetoed.

The California State Capitol in Sacramento
Image: Frank Schulenburg ( CC BY-SA 4.0 )
AI Existing / enacted California, United States

Official name: Transparency in Frontier Artificial Intelligence Act
Citation: Senate Bill 53 (Wiener), Chapter 138, Statutes of 2025; California Business and Professions Code sections 22757.10 and following
Year / status: Signed 29 September 2025. Core duties for the largest model developers effective 1 January 2026. Senate Bill 1047 (2024) was vetoed.

What it does

Senate Bill 53 requires large developers of frontier models, the statute’s term for the most advanced models, to publish a safety framework, assess catastrophic-risk claims in the statute’s sense, report critical safety incidents, and protect specified whistleblowers. It is a transparency and reporting law.

Where it applies / status

California law. It aims at large developers who meet the statute’s compute and revenue tests. The duties are state duties.

Documented benefits or intended effects

Intended effects: public safety-framework documents and a channel for incident reports to the state. As of 2026 it is early for an official evaluation of whether those filings reduced harm.

Caveats

Thresholds leave most developers out. A published framework is a public disclosure of the developer’s controls. Federal pre-emption fights are a live U.S. political risk. Senate Bill 1047 of 2024 was vetoed; this statute is Senate Bill 53.

Sources

California Legislature bill status and text; Governor’s 29 September 2025 signing release.

Related