Overview
- Gov. Gavin Newsom signed SB 53 on Sept. 29, targeting developers with more than $500 million in annual revenue that train “frontier” models exceeding 10^26 operations and tied to clearly defined catastrophic risk thresholds.
- Covered companies must publish a documented safety framework and post pre-deployment transparency reports detailing capabilities, intended uses, risk assessments, and mitigations.
- Developers must report critical safety incidents to California’s Office of Emergency Services within 15 days and alert an appropriate authority within 24 hours if imminent harm is identified.
- The California attorney general can enforce the law with civil penalties of up to $1 million per violation for noncompliance or false statements.
- The Act creates the CalCompute consortium to build a public cloud for safe AI research with a framework due by Jan. 1, 2027, as major AI firms voice cautious support and industry groups and the White House warn against a state-by-state patchwork.