Lawfare Warns Against Cybersecurity-Led Frontier AI Kill-Switch Powers
Emerging Technology Risk · 2 sources · Updated 2mo ago · First seen Apr 7, 2014

Lawfare warned that U.S. proposals to govern frontier AI models through emergency intervention or model “kill switch” authorities could place broad AI-safety decisions under a national-security framework without sufficient transparency or constraint. The article cites proposals from Reps. Ted Lieu, Nathaniel Moran, Jay Obernolte, and Lori Trahan, and Sen. John Kennedy, which would assign varying oversight or intervention roles to DHS, the Department of Commerce, private entities, and independent auditors.
The authors argue that treating AI governance primarily as a cybersecurity threat can foster secrecy, politicize oversight, and erode trust in cybersecurity institutions. They point to the reported Anthropic export-control episode involving the Fable and Mythos models as an example of alleged cyber risk driving expansive controls that were subsequently reversed, and call for narrowly scoped statutory powers, evidence-based decision-making, congressional disclosure, and independent review.


