California Governor Gavin Newsom on Friday issued an executive order that targets two hotly debated topics in Washington: independent verification organizations and so-called "kill switches." Newsom framed the move as filling a federal leadership gap, saying, "With Donald Trump and Congress asleep at the wheel, Governor Gavin Newsom is once again taking the lead to strengthen AI safety for all Americans."
Newsom first staked out a public position on AI in 2023 when he signed the state's first AI executive order, and his latest directive accelerates implementation of two laws signed last week. Both of those laws, which create paths for independent verification organizations, were endorsed by OpenAI and Anthropic.
Why this matters
In the absence of new federal legislation or broader action from the Trump administration, states are increasingly moving to regulate AI. According to the National Conference of State Legislatures, there are more than 600 AI-related bills across 47 states — a 24% increase from last year. California, Illinois and Massachusetts are among the most active states; all three are led by Democrats who are aiming to shape national policy on AI.
Political dynamics also play a role: politicians from both parties have at times leveraged opposition to data centers in campaigns, including GOP gubernatorial candidates in Pennsylvania and Wisconsin.
How the orders fit with industry efforts
The executive order aligns with some directions frontier AI labs are already pursuing: the concept of a kill switch, for example, corresponds to technical and safety work under way at major labs. The California order is intended to speed up the legal and administrative framework that enables independent verification and stricter safety enforcement.
Illinois’ recently passed bill requires companies to develop their own frameworks for frontier AI safety and mandates that a third party produce an annual report assessing whether developers comply with those requirements. The Massachusetts Senate bill goes further by calling for independent evaluations of the models themselves every four months; Anthropic has backed that approach, while OpenAI has supported a narrower version.
The broader political picture
Former Vice President Kamala Harris this week called for slowing advanced AI development, and Illinois Governor J.B. Pritzker — another presidential contender — promoted his AI bill on Friday, saying, "Trump says that the only AI guardrails that we need are a strong and smart president. And, of course, he is neither of those things."
Because federal action has lagged, states are acting faster to set rules — a development that is also shaping how major AI companies respond and adjust their public positions. OpenAI, for example, shifted its stance on California’s bills last week and may do so again in Massachusetts as concerns about AI rise.
What to watch next
Key items to monitor include whether OpenAI changes its position on the Massachusetts proposal and how states implement the new laws in practice. The evolution of state-level rules will be important both for industry compliance and for whether a coherent federal approach eventually emerges.
