California's SB 1047 — The Frontier AI Safety Bill a Governor Vetoed

SB 1047 was a California bill, introduced by State Senator Scott Wiener, that would have required developers of the most advanced large AI models to adopt safety measures. Its formal title was the Safe and Secure Innovation for Frontier Artificial Intelligence Models Act. It passed both houses of the legislature in August 2024, but Governor Gavin Newsom vetoed it on 29 September 2024, so it never became law. In his veto message the governor wrote that by focusing only on the most expensive and large-scale models, the bill "establishes a regulatory framework that could give the public a false sense of security about controlling this fast-moving technology."

A year later, on 29 September 2025, Newsom signed Wiener's successor bill, SB 53, the Transparency in Frontier Artificial Intelligence Act. The duty to take safety measures in SB 1047 became, in SB 53, a duty to publish safety frameworks and report incidents.

What the bill required

According to a September 2024 overview by the law firm Kilpatrick Townsend, the bill contained the following.

Wiener's office also listed CalCompute, a public computing cluster for startups, researchers and community groups, among the bill's pillars. According to the governor's veto message, the bill would also have established a state Board of Frontier Models to oversee development. The Assembly passed the bill 49 to 15 on 28 August 2024, and the Senate gave final approval the next day.

Who was for and against

SB 1047 made it plain that the AI industry itself was split over safety.

Anthropic's letter, as quoted by Reuters, said: "In our assessment, the new SB 1047 is substantially improved, to the point where we believe its benefits likely outweigh its costs. However, we are not certain of this". Pelosi's statement said: "The view of many of us in Congress is that SB 1047 is well-intentioned but ill informed," pointing to the bipartisan House task force on AI.

A few days after the veto Wiener told TechCrunch: "I've never had a bill with this level of misinformation," accusing opponents of running a propaganda campaign.

Why the governor vetoed it

Newsom's message does not reject AI regulation as such. It notes that California is home to 32 of the world's 50 leading AI companies and says the governor agrees with the author that the state cannot wait for a major catastrophe before acting. It then gives these reasons:

• The key question is whether the threshold for regulation should rest on the cost and computation needed to develop a model, or on a system's actual risks
• Focusing only on the most expensive, large-scale models could create a false sense of security; smaller, specialised models may prove equally or more dangerous
• The bill does not consider whether a system is deployed in high-risk environments, makes critical decisions or uses sensitive data, yet applies stringent standards to even basic functions of a large system
• Regulation must be informed by empirical analysis of AI systems and capabilities

The message adds that in the previous 30 days the governor had signed over a dozen bills regulating specific, known risks posed by AI. According to TechCrunch, Newsom set up a working group including Stanford's Fei-Fei Li at the same time.

The successor, SB 53

Building on the working group's report, published on 18 March 2025, Wiener introduced SB 53, which Newsom signed on 29 September 2025. According to the governor's announcement and an analysis by the Carnegie Endowment for International Peace:

Carnegie describes SB 53 as a lighter-touch approach than SB 1047, which had been criticised for creating new liability, requiring shutdown capabilities and imposing know-your-customer rules on cloud providers. A third-party audit provision in an earlier draft of SB 53 was dropped before passage. SB 53's core requirements took effect on 1 January 2026.

What it changed

The veto of SB 1047 and the enactment of SB 53 were the first US answer to the question how far a state can regulate the safety of frontier AI.

The position many opponents took — that AI should be regulated federally — later became federal policy. On 11 December 2025 President Trump signed Executive Order 14365, "Ensuring a National Policy Framework for Artificial Intelligence," directing the Attorney General to set up within 30 days a task force to challenge state AI laws and the Commerce Secretary to identify "onerous" state laws. The order's text does not name California, but it targets what it calls a patchwork of state regimes.

For workers, a notable thread runs from SB 1047 to SB 53: whistleblower protection. Protecting AI-company employees who speak up about risks that are not yet regulated echoes what employees demanded after the dissolution of OpenAI's Superalignment team.

SB 1047 marked the point at which "AI safety" moved from researchers' principles (the Asilomar AI Principles) and companies' voluntary pledges into the politics of statute text and a governor's signature. The bill that was vetoed became, a year later and in a different form, state law.