California legislature passes SB 1047

The California legislature passed SB 1047, the Safe and Secure Innovation for Frontier Artificial Intelligence Models Act, sending it to Governor Gavin Newsom’s desk. The Assembly approved the bill 48–16 on 28 August, and the Senate concurred with the Assembly’s amendments 30–9 the following day, clearing its final legislative hurdle roughly fourteen months after state senator Scott Wiener introduced it.

As passed, the bill would have required developers of models trained above a set computing-power and cost threshold to implement a written safety protocol, retain the ability to shut the model down, submit to third-party audits and report safety incidents, with the state attorney general empowered to sue over resulting harms. Wiener had already stripped out more contentious provisions during negotiation — including criminal perjury liability for false safety certifications and a proposed new state regulatory division — in an effort to secure enough votes and reduce industry opposition.

That opposition remained substantial through the final vote. OpenAI, Meta, Andreessen Horowitz and Y Combinator lobbied against the bill, and congressional Democrats including Nancy Pelosi and Zoe Lofgren, both representing Californian districts with large tech constituencies, publicly urged Newsom to veto it, arguing it would chill open-weight model releases and start-up formation. Supporters spanned an unusual coalition, from AI researchers Geoffrey Hinton and Yoshua Bengio to Elon Musk, who broke with much of the industry to back the bill.

The passage set up a decision squarely on Newsom’s desk: because most frontier AI developers were headquartered in California, the bill would have functioned as de facto national regulation regardless of federal inaction. Newsom vetoed the bill a month later, arguing it regulated model size rather than deployment risk, while announcing separate work on an alternative framework.