California enacted SB-53 in October 2025, establishing the first U.S. state-level frontier AI law. The law requires companies with >$500M annual revenue to publish safety frameworks and report catastrophic-risk incidents to state authorities when deploying models trained on more than 10^26 FLOPS. The regime prioritizes transparency and whistleblower protections while explicitly removing third-party audit mandates; it creates federal-state deference mechanisms and defines catastrophic risk narrowly (deaths, injuries, or >$1B economic harm from CBRN/cyber/evasion incidents). This establishes a binding state-level regulatory baseline that other jurisdictions and the federal government may use as a template, and creates asymmetric compliance cost and disclosure burden on the largest AI developers operating in California.
373 other tracked developments touching the same claims.