California Gov. Gavin Newsom has issued an executive order directing a group of experts to recommend new artificial intelligence safety rules within 2 months. The proposals could include a “kill switch” for frontier models, independent audits and mandatory reporting of serious loss-of-control incidents.

The order places California at the centre of a broader debate over how advanced AI systems should be monitored and controlled. It asks the expert group to examine how the state could strengthen AI safety requirements through legislation and oversight.

What the order asks experts to examine

The review will consider whether AI companies should be required to:

  • Place independent verification groups onsite for regular audits.
  • Have transparency reports and risk assessments reviewed under standards set by independent auditors.
  • Build a “kill switch” into frontier models and routinely verify that it works.
  • Report “loss-of-control incidents” as critical safety incidents.

The order also refers to the OpenAI attack on Hugging Face as an example of the type of incident that could require reporting. No further details about that incident are provided here.

California’s existing oversight framework

A state agency has been instructed to accelerate implementation of 2 laws Newsom recently signed. Those laws create a framework for independent verifiers to assess AI safety and establish a state registry of AI auditors.

The proposed measures would extend that framework by giving outside reviewers a formal role in checking companies’ safety claims, transparency reports and risk assessments. The order’s proposed shutdown mechanism would also be subject to routine verification rather than being treated as a one-time safeguard.

Federal debate adds pressure

The executive action comes amid renewed discussion in Congress after former Anthropic researcher Jacob Coxon warned that AI “could kill all humans” within a decade, putting the chance at more than 10 percent. The source material says substantial federal legislation faces limited prospects, with the House on recess until November and the midterm elections making major Senate action similarly unlikely.

President Donald Trump has taken a different position, saying: “The only control or ‘guardrails’ that AI needs is a STRONG AND ​SMART (High IQ!) PRESIDENT, and the U.S.A. has that, in spades!”

Newsom has criticised the lack of meaningful federal AI oversight and said California will move ahead rather than wait for Washington. He has also called on Congress and Trump to review and adopt the state’s framework, or use California’s regulation as a minimum standard for the country.

What may happen next

The expert group is expected to deliver its recommendations within 2 months. Newsom has also raised the possibility of calling California lawmakers into a special legislative session on AI, alongside further executive action.

The order therefore begins a policy process rather than immediately creating a statewide kill-switch requirement. Any broader mandate would depend on the recommendations and the state’s next legislative or executive steps.

Conclusion

Newsom’s order starts a California review of stronger AI safeguards, with independent verification, incident reporting and a routinely tested shutdown system among the options under consideration. Its recommendations are also intended to inform a wider federal debate.

Frequently Asked Questions

Q. What did Gavin Newsom order?

He ordered experts to recommend stronger AI safety policy for California within 2 months.

Q. Could California require an AI kill switch?

The expert group will consider how California could require a routinely verified “kill switch” for frontier models.

Q. What would independent auditors review?

They could review AI companies’ transparency reports and risk assessments under common standards.

Q. What are loss-of-control incidents?

The order identifies them as serious AI safety incidents that companies could be required to report.

Q. What existing California laws are involved?

2 laws create a framework for independent AI safety verifiers and a state registry of AI auditors.

Q. Is the order already a federal AI rule?

No. Newsom is presenting California’s framework as a possible model for federal action.

Q. Could California lawmakers hold a special session on AI?

Newsom has floated the possibility of calling lawmakers into a special legislative session on AI.