California SB 53, the Transparency in Frontier Artificial Intelligence Act, signed

California Governor Gavin Newsom signed Senate Bill 53, the Transparency in Frontier Artificial Intelligence Act (TFAIA), on 29 September 2025. The law puts transparency, safety incident reporting and whistleblower rules on developers of frontier AI models.

What it is

The governor's office tied SB 53 to recommendations in a state-commissioned expert report on frontier AI published earlier in 2025. It also framed the law as filling a gap left by the lack of comprehensive federal AI policy.

What it requires

According to the governor's announcement, the law:

• requires large frontier developers to publish on their websites a framework showing how they have incorporated national standards, international standards and industry-consensus best practices;

• creates a channel for frontier AI companies and the public to report potential critical safety incidents to the California Office of Emergency Services;

• protects whistleblowers who disclose significant health and safety risks from frontier models, backed by a civil penalty for noncompliance that the Attorney General enforces;

• sets up CalCompute, a consortium in the Government Operations Agency charged with a framework for a public computing cluster;

• tells the California Department of Technology to recommend updates to the law each year.

The announcement does not define which companies count as large frontier developers, so scope has to be checked against the statute itself.

What security leaders should do

• If your company trains frontier models, treat the published framework as a security document: regulators, customers and attackers will all read it, so describe controls without exposing sensitive detail.
• Build an internal route to classify and escalate potential critical safety incidents, so any report to the state is timely and consistent.
• Review whistleblower channels and non-retaliation policies with counsel.
• If you only consume frontier models, collect each vendor's published framework and use it in third-party risk reviews.