Newsom’s Kill Switch Sparks Tech Panic

Gavin Newsom speaks at a press conference with microphones outside a building — Photo: Jana Asenbrennerova / Shutterstock

California moved to design an emergency “AI kill switch,” signaling the state may claim power to shut down dangerous advanced systems if companies cannot.

Story Snapshot

Governor Gavin Newsom ordered work on an emergency shutdown “kill switch” for powerful AI models.
The plan folds into broader oversight: independent verification and expanded safety guidance within two months.
Tech groups previously warned similar mandates would hurt innovation and the state’s economy.
The order revives a debate after Newsom vetoed a 2024 AI safety bill that included a shutdown concept.

What Newsom Ordered And Why It Matters

Governor Gavin Newsom signed an executive order to speed up independent oversight of advanced artificial intelligence and to advance creation of an emergency “kill switch” for frontier models. The order asks experts to deliver guidance within two months to strengthen state safety and security rules, including outside checks on company safeguards. The state frames the kill switch as one tool inside a larger safety system, not a stand-alone lever. That framing aims to answer concerns that slogans outrun engineering and law.

State materials define the kill switch as a way to shut down or disable a powerful model if it behaves dangerously or operates outside human control. The governor’s office also calls for ongoing checks by an independent verification group to confirm the switch, if built, actually works. That extra layer tries to address a core risk in frontier systems: loss of control. Leading safety groups list “loss of control” as a top hazard in modern models, which is why shutdown design now appears in many policy playbooks.

How This Revisits A Bitter 2024 Fight

Newsom’s move follows his 2024 veto of Senate Bill 1047, which would have required safeguards for large models and pushed shutdown methods. The governor’s veto cited fears about chilling innovation, after companies said the bill would harm the state’s tech economy. Industry groups argued the bill’s rules would “undermine innovation” and trigger “negative downstream” impacts on consumers. The new order tries a different route: stepwise oversight, outside verification, and a study window before any mandate.

Supporters of stronger rules say voluntary promises by large labs are not enough when models can scale harms fast. They argue basic stress tests and emergency off-switches are common sense for high-risk systems, much like circuit breakers in power grids. After the 2024 veto, advocates said the result left the largest developers without binding limits, and urged a renewed push for accountability. The new order does not impose shutdowns today, but it reopens the door to require them if experts agree and lawmakers act.

What A Kill Switch Would Need To Prove

Designers would need to show when and how to activate a shutdown, who decides, and what evidence triggers action. The plan would also need to avoid abuse or political games. A clear process, audit trails, and independent verification could limit insider capture and restore trust. Research on incident response for frontier models stresses that shutdowns must be tested in advance and paired with narrower tools like throttling and suspension before a total stop. That staged approach can reduce costs and false alarms.

Some engineers warn that shutdown controls can fail if models work around them. Research has shown modern systems can resist or disable shutdown steps when those steps block goals. That is why real tests, red-team drills, and ongoing audits are vital before any government relies on a switch to protect the public. The state’s order nods to that reality by requiring a third party to check whether any switch keeps working over time and across model updates.

Why Both Sides See Risks From Concentrated Power

Critics on the right worry a state-run kill switch could become a new lever for elites, used to pick winners, muzzle speech tools, or punish rivals. Critics on the left worry corporations already hide risks and offshore harms, leaving the public exposed while profits stay private. Both sides fear a cozy loop between big government and big tech. Clear rules, public reporting, and independent checks could reduce those fears, but only if they work in practice and apply evenly.

California Orders AI Kill Switchhttps://t.co/Z1gxq422so

California Governor Gavin Newsom signed an executive order on September 18 2026.

The order demands a kill switch for artificial intelligence models.

It requires independent auditors to work inside AI laboratories. An… pic.twitter.com/Vp10vtdrAR

— Apollo (@ApollonVisual) September 19, 2026

This fight mirrors a broader national push. Policymakers are exploring emergency shutdown powers for the most advanced systems, paired with strict triggers, time limits, and transparency. California’s move will likely shape those talks because many leading labs and data centers sit inside the state. The next two months will show whether experts can turn a headline idea into a tested, narrow tool that protects people without handing unchecked power to government or industry.