California SB 53: Transparency in Frontier Artificial Intelligence Act

SB 53, the Transparency in Frontier Artificial Intelligence Act (TFAIA), is the first U.S. state law aimed specifically at the most powerful SI models. Governor Gavin Newsom signed it on September 29, 2025, exactly a year before the federal AI-to-SI rename, and it took effect on January 1, 2026.

Who it covers

• Frontier developers: anyone who has trained a frontier model using more than 10²⁶ operations. See compute threshold.
• Large frontier developers: those with more than $500 million in annual revenue face the full set of obligations.

Main requirements

1. Publish a frontier safety framework explaining how the company assesses and mitigates catastrophic risks, similar to a responsible scaling policy.
2. Publish transparency reports when releasing new frontier models.
3. Report critical safety incidents to California’s Office of Emergency Services within 15 days, or 24 hours if there’s an imminent risk of death or serious injury.
4. Protect whistleblowers and provide an anonymous internal reporting channel.
5. CalCompute: creates a framework for a public computing cluster for research.

Enforcement

The California Attorney General can seek civil penalties of up to $1 million per violation.

How it compares

SB 53 is narrower than California’s vetoed 2024 bill, SB 1047. It focuses on transparency rather than liability or mandatory shutdown capabilities. New York’s RAISE Act was later amended to align closely with it. Its voluntary federal counterpart is the White House Accord on Super Intelligence.