Anthropic endorses California’s AI safety bill, SB 53 (techcrunch.com)

Anthropic has officially endorsed California’s Senate Bill 53 (SB 53), a pioneering AI safety bill that would impose mandatory transparency and safety reporting requirements on the largest AI model developers, including OpenAI, Google, and xAI. The bill targets “frontier AI” systems capable of causing catastrophic risks—defined as events causing significant loss of life or damages exceeding a billion dollars—by requiring companies to develop safety frameworks and publicly disclose safety and security reports before deployment. It also introduces whistleblower protections for employees raising safety concerns. Anthropic’s support signals a rare alignment with regulators amid strong industry resistance from groups like the CTA and tech investors fearing state-level rules could stifle innovation. SB 53 focuses on mitigating extreme risks such as the use of AI to facilitate biological weapons or cyberattacks, deliberately steering clear of more immediate issues like AI-generated misinformation. While Anthropic favors federal standards for AI governance, it acknowledges that legislative action cannot wait for Washington consensus. The bill has been refined to remove contentious third-party audits, reflecting a pragmatic balance between oversight and industry concerns. If passed, SB 53 would set a legal baseline for AI safety practices that many leading labs currently undertake voluntarily but are not yet obligated to maintain by law. Experts see SB 53 as a measured, technically informed step toward proactive AI governance in the U.S., especially as federal regulation remains uncertain.