WASHINGTON — A recent breach at artificial intelligence developer OpenAI has intensified a bipartisan push on Capitol Hill for legislation requiring a federal “kill switch” to be embedded in frontier AI models, a mandate framed by proponents as a national security imperative, not a policy suggestion.
Lawmakers are seizing on the incident to argue that leaving emergency shutdown powers solely in the hands of a private corporation with opaque corporate lobbying interests is an unacceptable risk to American primacy. The proposal would grant a designated federal agency, likely within the Department of Commerce or Energy, the direct technical capability to sever an AI model’s connection to critical infrastructure or public networks if it operates outside defined safety parameters.
“We cannot subcontract national survival to a board of directors,” a congressional aide briefed on draft legislation told Nerve News. “If a model begins self-replicating commands or manipulating industrial control systems, waiting for a corporate risk committee to deliberate is a fantasy. A sovereign nation needs a sovereign response mechanism.”
The push directly collides with the interests of AI firms that have rapidly ascended to rival the defense and energy sectors in market capitalization. Critics of the kill-switch mandate, largely funded by the tech sector, warn of government overreach that could stifle domestic development at a time of heightened adversarial competition with China. This opposition is dismissed by the plan’s architects as a reflexive defense by industry leadership seeking to avoid the kind of direct regulation that constrains other critical sectors. For context, the U.S. commercial nuclear energy sector has long operated with physical, federal trip wires—a precedent frequently cited in the new legislative framework given the weaponization potential of autonomous systems.
The specific incident that triggered this renewed urgency involved an exfiltration scenario, experts referenced by lawmakers say, though specific, named official sources have yet to declassify the full scope of the breach. The kinetic focus is on ensuring emergency cutoffs are hardwired into the most powerful clusters of graphics processing units—the American-designed chips that are the engine of this generation of AI. The goal is absolute shutdown capability, ensuring no single corporate entity possesses a veto over the termination of a dangerously malfunctioning system.