ComplianceEffective date1 Jan 2026US
California requires frontier AI developers to publish safety frameworks and report incidents
In force now
Since 1 Jan 2026, developers that train models above 10^26 operations publish a transparency report with each new model, and those with more than $500 million in annual revenue publish a framework for catastrophic risk. Critical safety incidents go to the state Office of Emergency Services within 15 days, and the law protects whistleblowers. The Attorney General can seek up to $1 million per violation.
Document: If you train or fine-tune at frontier scale, check each release against your published framework and that incidents reach the state within 15 days.
Who it reaches
It reaches generative AI services with users in the US, at any size. Developers training models above 10^26 operations; the framework duty applies above $500 million in revenue.
Lands on Policy & Standards, Legal & Regulatory and Engineering.
Learn more
Read the sourceFuture of Privacy Forum
Not legal advice. Read the source.