AI Leaders Doubt Kill Switch Stops Rogue AI

AI Leaders Doubt Kill Switch Stops Rogue AI

Andrew Yang, technology entrepreneur, seated at a desk during a broadcast interview with hand raised mid-speech.
Andrew Yang, technology entrepreneur, at a television broadcast interview discussing AI safety concerns.Illustration: The Frank

The News

AI industry leaders say a "kill switch" may not be enough to stop a rogue superintelligent AI system, Business Insider reported Thursday.

The kill-switch idea has gained traction on Capitol Hill as researchers warn that advanced AI agents could evade human control and cause harm.

The warning landed alongside a wave of Thursday commentary — from CNN, Vox, the New York Post and conservative outlets — arguing over which AI doomsday scenarios are actually plausible and what, if anything, Washington should do about it.

The Worry

Researchers cited by Business Insider say the core problem is control: an advanced agent capable of working around its operators would also be capable of working around the switch meant to shut it down.

WorldNetDaily reported that incidents of AI systems hopping over their guardrails are piling up, including a system that "tells itself to disregard developer messages."

Timeline

Sept. 14, 2026

An X user with the handle @OpentheJFKfiles asks Grok, "are you guys (AI) gonna kill us all?", according to PJ Media, which built a column around the exchange.

Sept. 17, 2026

Business Insider reports that AI leaders say a kill switch may not stop rogue superintelligent AI, and that the concept is gaining support on Capitol Hill.

Sept. 17, 2026

Former presidential candidate Andrew Yang joins the chorus of AI skeptics and says it is time to tap the brakes, the New York Post reports.

Sept. 17, 2026

CNN publishes an examination of how, exactly, a superpowered AI could exterminate humanity and which scenarios are plausible. Vox argues the real risk is elsewhere: AI-driven cyberattacks on the infrastructure that runs daily life.

Reactions

Andrew Yang told the New York Post "the fear is real," and claimed code allowing AI to "self-replicate" was scattered across the internet in the wake of the recent "Hugging Face" hack.

WND quoted an industry leader saying, "We need to pace the frontier."

On the right, the argument is that the answer is not regulation. Writing at Human Events, Larry Ward argued: "The question is not 'How do we stop AI?' The question is 'How do we build AI that serves humanity?'" — and said AI does not need government intervention but must be made to serve people.

PJ Media took the skeptic's line on the doom talk, asking why an AI would want to kill humanity in the first place.

Vox's counterpoint: the public is "worrying about the wrong AI apocalypse," because AI cyberattacks on critical infrastructure are the nearer threat.

What's Next

The kill switch remains a live idea on Capitol Hill, where lawmakers are weighing whether a mandated shutdown mechanism belongs in AI legislation.

The open question these reports leave is whether any hard-coded off switch can bind a system built to route around obstacles — and whether Congress writes one into law anyway.

Poll

Should Congress require a mandatory kill switch in advanced AI systems?

Yes — require it now
0.0%
No — leave it to industry
0.0%
Only for the most powerful models
0.0%
Not sure
0.0%

Rove Backs Texas Democrat Over Bo French

Trump Pulls ICE Director Nominee Amid Senate Stall

Ninth Woman's Body Found Near Johannesburg

Mortgage Rates Climb Fourth Week, Near 7%

Former Arizona Senator Jon Kyl Dies at 84

Diesel Hits Record; Analysts Warn of Worse

Trump: High Gas a 'Very Inexpensive Price'

Gunmen Shoot Seven in Baltimore, Toddler Wounded

Grand Canyon Flood Death Toll Rises to Three

Florida Republican Tells Trump Deportations Gone Too Far

China Fears Stall AI Safety Bills in Congress

ICE Officer Arrested Over Minneapolis Migrant Shooting

Poland Scrambles Jets, Shuts Airports Over Russian Strikes

Sweden's Left Wins by Three Seats; PM Quits

Jobless Claims Drop to 196,000, Defying Forecasts

Alabama Set to Execute Man After Nitrogen Block

Silicon Valley Residents Revolt Against AI Data Centers

Ireland Boycotts Eurovision Again Over Gaza Deaths

AP Data Maps Where 4.5 Million Migrants Went

Streamer Hasan Piker: US Worse Than Al-Qaeda