California’s attempt to impose new safety duties on large artificial intelligence developers has been stopped by Governor Gavin Newsom, who vetoed SB-1047 after months of pressure from supporters and critics across the AI industry.
The bill would have required makers of large AI models to use safety tests and kill switches intended to prevent potential “critical harms.” Newsom said the goal was understandable, but the method was too broad in some places and too narrow in others.
Why Newsom rejected SB-1047
Newsom’s main objection was that SB-1047 tied its regulatory framework to the scale and expense of AI models. In his view, that approach could make the public feel protected while leaving other meaningful risks outside the bill’s reach.
“By focusing only on the most expensive and large-scale models, SB-1047 establishes a regulatory framework that could give the public a false sense of security about controlling this fast-moving technology,” Newsom wrote.
He also warned that “Smaller, specialized models may emerge as equally or even more dangerous than the models targeted by SB-1047—at the potential expense of curtailing the very innovation that fuels advancement in favor of the public good.”
That argument goes to the central problem facing AI regulation: danger may not always map cleanly to model size. A large system used for a simple task may pose one kind of risk, while a smaller system deployed in a sensitive setting may pose another.
Newsom said SB-1047 did not properly consider whether an AI system was used in high-risk environments, involved critical decision-making, or handled sensitive data. Instead, he said, it would apply strict standards to basic functions if they were deployed by a large system.
“I do not believe this is the best approach to protecting the public from real threats posed by the technology.”
The risks California still wants to confront
The veto does not mean Newsom dismissed AI risk. His statement listed several areas where he said rapidly evolving risks could be addressed in a more targeted way.
Those risks include “threats to our democratic process, the spread of misinformation and deepfakes, risks to online privacy, threats to critical infrastructure, and disruptions in the workforce.”
California already has a number of AI laws on the books that target some potential harms, and many other states have signed similar laws. The veto therefore leaves the state with existing rules, but without SB-1047’s broader framework for large AI model developers.
The distinction matters. Newsom’s position suggests that AI safety rules should focus more on deployment, data sensitivity, and real-world use than on the technical size of a model alone.
Supporters saw the bill as overdue oversight
State Senator Scott Wiener, who co-authored SB-1047, sharply criticized the veto. In a social media post, he called it “a setback for everyone who believes in oversight of massive corporations that are making critical decisions that affect the safety and welfare of the public and the future of the planet.”
Wiener argued that voluntary safety promises from AI companies are not enough. He said the lack of effective government regulation means “we are all less safe as a result” of the veto.
SB-1047 had support from major figures in the AI field, including Geoffrey Hinton and Yoshua Bengio. xAI leader Elon Musk also advocated for the bill’s passage, saying it was “a tough call” but that AI needs to be regulated “just as we regulate any product/technology that is a potential risk to the public.”
SAG-AFTRA also supported the bill. The actor’s union called it a “first step” toward protection against known dangers such as deepfakes and nonconsensual use of voice and likeness for its members.
Opponents warned about costs and open models
Critics argued that SB-1047 took a heavy-handed approach. Some worried about the legal liability it could create for open-weight models when others used those models for harmful purposes.
After the bill passed the state Assembly in August, a group of California business leaders sent an open letter to Newsom urging a veto. They described the bill as “fundamentally flawed,” saying it “regulates model development instead of misuse” and would “introduce burdensome compliance costs.”
Lobbyists for major tech companies including Google and Meta publicly opposed SB-1047. At the same time, a group of employees from those and other large technology companies supported its passage, showing that the divide was not simply between companies and outsiders.
OpenAI Chief Strategy Officer Jason Kwon also publicly urged a veto. In an open letter, he argued that federal regulation would be more appropriate and effective than “a patchwork of state laws.” According to the source article, early efforts to build federal legislation have stalled in Congress amid policy road maps and working group reports.
What the veto leaves unresolved
The SB-1047 debate exposed a larger question: how should governments regulate AI when the technology changes quickly and risks can appear in many different forms?
Newsom had already signaled concern about the bill’s reach. At the 2024 Dreamforce conference earlier this month, he spoke about “the sort of outsized impact that legislation [like SB-1047] could have, and the chilling effect, particularly in the open source community… I can’t solve for everything. What can we solve for?”
That question remains open after the veto. Supporters of SB-1047 argue that powerful AI systems need enforceable oversight. Opponents argue that rules should be aimed at misuse, sensitive deployments, and concrete harms rather than model development itself.
For now, California has stepped back from this particular AI safety framework. But the concerns that drove SB-1047, from deepfakes and misinformation to privacy and critical infrastructure, remain part of the state’s regulatory agenda.