Newsom orders California toward an AI "kill switch," backed by embedded auditors
A new executive order pushes past disclosure and toward verification -- directing state agencies to get independent evaluators physically inside frontier AI labs, and to fast-track two bills already aimed at the same goal.
California Governor Gavin Newsom signed an executive order on September 18 directing the state's Government Operations Agency to convene national experts and deliver recommendations, within two months, on accelerating independent oversight of frontier AI models -- including the groundwork for an emergency shutoff mechanism, verified on an ongoing basis by independent organizations, that could be triggered if a model started operating outside acceptable bounds. "We're not waiting to act -- we're going to speed up our work on substantial and responsible AI oversight before it's too late," Newsom said.
From "tell us what happened" to "let us watch"
The order's most concrete step is a shift from after-the-fact disclosure to standing access: it calls for frontier AI companies to host designated independent verification organizations onsite in their own labs for regular audits, and for those same organizations to verify companies' safety frameworks, transparency reports, and risk assessments rather than take them at face value. It also updates the state's definition of a "critical safety incident" to explicitly include loss-of-control incidents -- a category the order ties to a referenced attack involving Hugging Face -- and directs agencies to accelerate implementation of two bills already moving through Sacramento, SB 813 and AB 1405, both aimed at building out an AI auditor registry. The order closes by calling on Congress and the President to adopt California's framework as a national baseline, not just a state one.
Why it matters for anyone buying AI tools, not just building them
An executive order isn't a law, and "convene experts and report back in two months" is a process step, not a finished regime -- the actual kill-switch mechanism, and who gets to pull it, remains to be defined. But the direction is unambiguous: California is moving from asking AI labs to describe their own safety posture toward requiring someone independent to verify it in person, on an ongoing basis. For a company evaluating which AI vendors to trust with real spend, that's the same distinction Merit AC's own tracker is built around -- self-reported claims of value and safety are a starting point, not a substitute for someone checking the work.