California Passes AI Safety Bill

California lawmakers have approved SB 53, a bill designed to increase transparency and safety in AI development. The bill mandates that large AI companies publish safety frameworks, disclose risk assessments, and report critical incidents. It also provides whistleblower protections for employees and contractors. SB 53 applies to AI developers with revenue exceeding $500 million and models trained with substantial computing power. Penalties for violations range from $10,000 to $10 million.

The bill establishes CalCompute, a state-run public cloud cluster within the University of California system, to promote safe and equitable AI research. It requires companies to disclose safety incidents involving dangerous deceptive behaviour by autonomous AI systems. The legislation focuses on transparency rather than liability, making voluntary safety commitments legally enforceable.

SB 53 now awaits Governor Newsom's consideration, who previously vetoed a similar bill. If signed, it would enact world-leading AI safety disclosure requirements. The bill has garnered support for its balanced approach to AI risk and innovation.