Particle.news
Download on the App Store

California Finalizes Frontier AI Safety Law Ahead of 2026 Rollout

The law imposes public risk disclosures for top AI labs with confidential incident reporting to California emergency officials.

Overview

  • SB-53, signed by Governor Gavin Newsom on September 29, 2025, establishes new transparency, safety, and whistleblower protections for frontier AI systems.
  • The statute takes effect January 1, 2026 and applies to developers training models above 10^26 compute or earning more than $500 million in annual revenue.
  • Covered firms must publish a frontier AI framework, conduct catastrophic-risk assessments before deployment, and post model-specific transparency reports.
  • Developers must submit quarterly summaries of internal risk assessments to Cal OES, report critical safety incidents within 15 days, or within 24 hours if there is imminent risk of death or serious injury, with reports kept confidential.
  • The law authorizes enforcement by the California Attorney General with penalties up to $1 million per violation and directs the state to build a CalCompute public cloud consortium for safer AI research.