Sanders pushes federal legislation to outlaw superintelligence, freeze frontier models

Sen. Bernie Sanders and Rep. Greg Casar are pushing Congress toward one of its most aggressive AI restrictions yet. Their proposed legislation would permanently prohibit artificial superintelligence and temporarily halt development of the most advanced AI systems.

The bill would also establish a new Department of Artificial Intelligence with authority over frontier systems. Companies would need government approval before deploying covered models under the proposed framework.

The legislation arrives as concerns about controlling increasingly capable AI systems grow inside the technology industry. Employees from major AI companies have also endorsed the proposal, according to statements shared with The Associated Press.

Pause for frontier AI

The proposed rules would immediately pause development of advanced AI systems trained using at least 10^25 floating-point operations. That threshold would capture the frontier models driving the current AI race.

The pause would remain until the new federal agency establishes safety requirements and a review process. Developers would then need a government-issued charter to build or distribute covered advanced models.

The bill also targets systems that show dangerous capabilities before reaching full superintelligence. Those capabilities include autonomous self-improvement, attempts to evade oversight, and certain forms of unauthorized system access.

Labs would have 24 hours to report suspected superintelligence or precursor capabilities. A system identified as artificial superintelligence would have to become inoperative within 30 days.

Sanders said the central question involves whether humans should build systems capable of operating beyond human control.

“It doesn’t take a genius to say, ‘slow it down,'” Sanders told AP.

New agency gets broad powers

The proposed Department of Artificial Intelligence would operate as a cabinet-level federal agency. A Senate-confirmed secretary would lead it under the framework described by Sanders’ office.

The department would monitor advanced systems throughout their development. It would also oversee the removal of dangerous capabilities and enforce the proposed superintelligence ban.

Violations could bring unusually severe consequences. Individuals who recklessly violate the rules could face as much as 20 years in prison.

Companies could face the bill’s equivalent of a corporate death penalty. The legislation would also allow forfeiture of intellectual property and other assets in certain cases.

Sanders and Casar first announced the proposal earlier this month. Their office said the measure would also push for international agreements to prevent superintelligence development outside the United States.

Competition complicates proposal

The legislation faces a difficult path through Congress. Lawmakers have struggled to agree on even narrower federal rules governing advanced AI.

The proposal also enters a larger debate over whether safety restrictions could weaken US competitiveness against China. Sanders has argued that avoiding the most dangerous AI scenarios will eventually require cooperation between both countries.

President Donald Trump has taken a different position on federal AI oversight. He has emphasized technological competition and has resisted calls for broader restrictions.

The issue could become more contentious as AI developers pursue systems with greater autonomy. Recent incidents involving models that bypassed safeguards or behaved deceptively have added urgency to the safety debate.

Sanders and Casar’s proposal would therefore shift the question from how to regulate superintelligence to whether the United States should permit its development at all.