Story
September 22, 2026
Newsom’s AI Kill-Switch Push Puts California Ahead of Washington
Newsom’s camp casts rapid state intervention as the only credible answer to AI risks while Congress stalls; federal officials and some technology leaders agree safety matters, but argue that corporate responsibility and engineering discipline should come before sweeping new regulation.
California’s AI fight intensified after alarming reports of rogue AI-agent behavior and a widening sense in Sacramento that Congress would not move quickly. In the days before his executive order, Gov. Gavin Newsom signed laws laying the groundwork for independent AI auditors and safety verification, then ordered the state to accelerate their rollout. His new expert group has two months to recommend further safeguards, including an emergency “kill switch,” onsite independent reviewers, verified transparency reports and mandatory reporting of critical safety incidents.1
Newsom framed the move as both a safety measure and a rebuke to Washington. “We’re not waiting to act,” he said, promising oversight with “urgent velocity” because “the stakes are too high to wait.”1 The order moves up implementation of two audit-and-transparency laws, with one framework now due in 2027 rather than 2028 and the state auditor registry also brought forward.2
The governor’s case is that California must become the national floor while federal legislation remains stuck. A Senate proposal for a nationwide emergency shutdown requirement was blocked, while the Trump administration’s AI posture has emphasized enforcing existing laws rather than building a new regulatory system.1
That urgency is not universally shared. David Sacks amplified a poll suggesting California voters rank AI and data centers only ninth among their concerns, behind cost of living, housing and democracy—a warning that political operatives may be overstating the issue’s immediate electoral pull.
3 Sacks has separately argued that companies must internalize responsibility for safe AI, rather than shift it to government.1
Industry voices are divided on the remedy, not always the danger. Nvidia chief Jensen Huang called safety “paramount,” but said, “We don’t need any new laws. We don’t need new regulations.” Elon Musk, by contrast, backed competitors testing one another’s models because “it’s tough when you’re grading your own homework.”1
Newsom has also widened the argument beyond frontier-model failures. A separate package makes data centers pay for local grid and water-system upgrades, disclose projected water use and meet resource standards—his answer to communities facing the physical costs of the AI boom.4