What happened. Governor Gavin Newsom signed SB 53, authored by Senator Scott Wiener, a year after vetoing the same author's broader SB 1047. The governor's office summarised the law as follows: large frontier developers must publish on their websites a framework describing how they incorporate national, international and industry-consensus standards; a mechanism is created for developers and the public to report potential critical safety incidents to the Office of Emergency Services; whistleblowers who disclose significant health and safety risks are protected; a consortium within the Government Operations Agency is to design a public computing cluster, CalCompute; the Department of Technology must recommend annual updates; and noncompliance carries civil penalties enforced by the Attorney General.
Scope. The statute targets developers of 'frontier models', defined as foundation models trained using more than 10^26 integer or floating-point operations (including fine-tuning and other later modifications), and imposes the fuller duties on 'large' frontier developers whose annual gross revenue, together with affiliates, exceeded $500 million in the preceding calendar year (Business and Professions Code § 22757.11(i)-(j)). Most companies using or fine-tuning AI are therefore outside its scope.
Why it matters. It is the first US state law aimed specifically at the developers of the most capable models, and it chooses disclosure and incident reporting over licensing or liability for harms. The co-authors of a report by experts convened by the governor said in the governor's announcement that the law moves towards the 'trust but verify' principles of their report. Supporters see a light, evidence-building first step; critics argue either that it is too weak to change behaviour or that a single state should not set rules for a national industry.
Federal tension. In December 2025 the President signed Executive Order 14365, directing federal agencies to challenge state AI laws deemed onerous and asking for a federal framework that would preempt them. Whether SB 53 is targeted, and whether such challenges succeed, will shape how much room sub-national governments have to regulate frontier AI.
What to watch. The quality and comparability of published frameworks; the first incident reports and how the state handles them; enforcement by the Attorney General; and litigation or federal legislation on preemption.