Anthropic endorses California's revised SB 53

Anthropic published a formal endorsement of SB 53, the California transparency bill authored by state senator Scott Wiener that Governor Gavin Newsom signed into law later that month. The company said the revised bill took a “trust but verify” approach — requiring covered developers to publish safety frameworks and report critical incidents — rather than the prescriptive technical mandates, including a shutdown-capability requirement and third-party audits, that had characterised its predecessor, SB 1047.

Anthropic framed its support as consistent with a standing position that it had “long advocated for thoughtful AI regulation” and that the bill largely codified practices the company and other frontier labs had already adopted on a voluntary basis through their own published safety frameworks. It argued SB 53 created “a level playing field where disclosure is mandatory, not optional,” rather than leaving compliance dependent on individual companies’ goodwill.

The endorsement was notable because Anthropic had broken from the rest of the industry the year before in supporting SB 1047’s amended language even as OpenAI, Meta and much of Silicon Valley lobbied against it; Newsom vetoed that bill regardless. With SB 53, Anthropic’s public backing came alongside — rather than against — the wider industry consensus, since the narrower bill imposed disclosure obligations without the liability and audit provisions labs had most strongly opposed. The endorsement functioned as a signal to lawmakers, ahead of the bill’s final vote, that at least one major frontier developer regarded transparency-only regulation as workable, at a moment when a competing federal effort to preempt state AI legislation was also under discussion.