California Governor Gavin Newsom signed an executive order directing state agencies to develop new artificial intelligence safety plans — and to study whether the state should have the power to shut down an AI system, an idea commonly called a "kill switch."
The order marks a sharp turn in how the country's largest tech state regulates AI. Last year, California lawmakers passed a stricter AI safety bill. Newsom vetoed it, arguing the measure was too broad. The new executive order instead asks agencies to build safety rules from inside the government.
Under the order, state agencies are tasked with drafting AI safety plans, reviewing how the technology is used in state operations, and evaluating safeguards for high-risk systems. The "kill switch" language is the part drawing the most attention — it signals the state is considering whether officials should be able to suspend or shut down an AI model that is judged dangerous.
Supporters describe the move as a practical middle path: real oversight without a single sweeping law that could chill an industry worth hundreds of billions of dollars and centered in the Bay Area. Critics say an executive order can be reversed by the next governor, and that it substitutes agency guidance for binding rules.
The order also lands in a national vacuum. Washington has not passed comprehensive AI legislation, leaving states to act on their own. That has pushed the debate into the 2028 campaign, where Democrats including Newsom and Illinois Governor J.B. Pritzker have leaned into AI safety as a message.
What the order does not settle is the central question: what would trigger a shutdown, and who would hold the authority to order one. Those details are left to the agencies, with timelines still taking shape. How California writes those rules could set the standard other states copy.