California Governor Newsom Signs Executive Order to Advance AI Safety Regulation
nashnova research
California Governor Gavin Newsom signed an executive order directing experts to draft AI safety guidelines within two months — including mandatory third-party safety audits and a kill switch requirement — the most concrete state-level move yet as federal AI legislation remains stalled.
What does this executive order actually require?
Newsom ordered a panel of experts to deliver concrete guidance within two months, aimed at strengthening California's existing AI safety legal framework.
Two hard requirements are on the table: independent third parties must write safety plans for frontier AI companies, and firms must build emergency shut-off switches — mechanisms to take a model offline in a crisis.
This means → if the proposals become law, companies headquartered in California — OpenAI, Anthropic and others — face a new compliance layer, with safety oversight shifting from self-assessment to external review.
Why is California moving first?
Newsom was blunt in his statement: the federal government has "utterly failed" to establish meaningful AI regulation — even as AI company CEOs themselves call for oversight.
The backdrop: an Anthropic researcher publicly warned this month that AI poses a risk of "ending humanity," then resigned. The episode drew attention in Washington — yet according to CNBC, Congress remains unlikely to pass substantive AI legislation before the 2026 midterms.
In plain terms = no one in Washington wants to touch this in an election cycle. President Trump has dismissed AI safety concerns as a "hoax." So California decided to go it alone.
What cards does California already hold?
Newsom cited a law he signed earlier this month that established an operational framework for third-party AI auditors.
He used that as the basis for claiming California has built "the nation's strongest AI regulatory framework" and argued the state's policies should serve as the national benchmark.
This reflects a broader strategy: California is not just regulating locally — it aims to force the federal government's hand through first-mover legislation, much as it did with automobile emission standards.
Who else in the Democratic Party is following?
Pennsylvania Governor Josh Shapiro called for "third-party" oversight of frontier models — aligned with Newsom's direction.
Arizona Senator Mark Kelly said he would introduce legislation creating a fund — financed by AI companies — to support workers displaced by AI.
New Jersey Senator Cory Booker went furthest, urging Trump to convene a special session of Congress dedicated to AI risks. This means → AI regulation is becoming a competitive issue in the 2028 Democratic presidential primary — all three are viewed as potential candidates.
What does this mean for the market?
The executive order itself carries no direct legal force. The real impact depends on what the expert panel delivers in two months and on the legislative process that follows.
Newsom's term expires on January 4, 2027. He has not declared a presidential run but says he is considering one. This means → the order is both a policy action and a political signal, building a "responsible regulation" narrative for a possible campaign.
In plain terms = frontier AI companies will not face fines or shutdowns tomorrow, but California's direction is clear: the compliance bar will only rise, especially on third-party safety audits and emergency shutdown mechanisms.
市场有风险，内容仅供研究参考，不构成投资建议。