Trump calls for AI 'kill switch'

President Donald Trump said there “should be” a government artificial intelligence “kill switch” in an interview about risks to the banking system. (newsmax.com) The comment aired April 15 on Fox Business Network’s “Mornings with Maria,” in an interview taped April 14, after Maria Bartiromo asked whether artificial intelligence could undermine confidence in banks. Trump said it “probably” could, but also said the technology could make banking “better and safer and more secure.” (newsmax.com) The immediate backdrop was Anthropic’s April 7 release of Claude Mythos Preview, a model the company said showed unusually strong cybersecurity capabilities and would not be made generally available. Anthropic’s own safety write-up said the system could materially raise offensive cyber risk. (red.anthropic.com) Banks have been on alert since then. Reuters reported this week that European Central Bank supervisors planned to question bankers about whether Mythos could accelerate cyberattacks against financial firms that still run on older software. (aol.com) Trump’s remark also lands in the middle of a broader White House push that has mostly emphasized faster deployment, lighter regulation and American dominance in artificial intelligence. A White House fact sheet on April 7, 2025 said the administration was replacing what it called a “risk-averse approach” with a “forward-leaning, pro-innovation” one for federal use and procurement. (whitehouse.gov) That agenda expanded on July 23, 2025, when the White House released “America’s AI Action Plan,” a package of more than 90 federal policy actions focused on innovation, infrastructure and international security. The plan also called for procurement rules favoring frontier model developers whose systems are “objective and free from top-down ideological bias.” (whitehouse.gov) On March 20, 2026, the White House followed with a national artificial intelligence legislative framework that called for stronger federal leadership, protections for children, anti-scam measures and action on national security risks. The document did not spell out any technical “kill switch” mechanism for private models. (whitehouse.gov) The phrase “kill switch” can mean very different things in practice: cutting a model off from users, shutting down connected tools, or blocking specific high-risk functions. Trump did not describe which of those powers he meant, who would hold them, or what legal trigger would activate them. (newsmax.com) Anthropic has taken a narrower route with Mythos itself. The company said the model would stay limited to a small set of partners and that it was withholding general release because of cybersecurity concerns. (red.anthropic.com) For now, the clearest new fact is the split inside Trump’s own artificial intelligence message: speed up deployment, cut rules, and keep a government stop button in reserve. (whitehouse.gov)