The warning came first: Sanders has argued that the race toward ever more capable AI needs to slow down, not simply be managed at the margins. His new proposal with Rep. Greg Casar is built on that premise, seeking to prohibit the development of artificial superintelligence.¹

The lawmakers then put legislative muscle behind the rhetoric. Their Ban Artificial Superintelligence Act would bar the creation of systems described as capable of the “destruction or disempowerment of humanity,” including by overthrowing the government.² It would also pause development of certain advanced, frontier AI models while Washington builds the machinery to oversee them.

At the center of the plan is a proposed Department of Artificial Intelligence, staffed by scientists and charged with monitoring frontier models, stripping dangerous capabilities and supervising the destruction of prohibited superintelligent systems. AI companies would need the department’s permission before developing advanced models.³

The enforcement mechanism is deliberately severe. Company leaders — as well as unaffiliated “rogue actors” — could face up to 20 years in prison for violations, a penalty the bill compares with existing sanctions for unlawful nuclear-weapons development.²

Sanders frames the legislation as a rejection of self-regulation by the firms driving the AI boom. “When you are racing towards a cliff, you don’t just ease up on the gas pedal. You hit the brakes,” he said, adding that “we cannot let a handful of Big Tech CEOs write their own rules.”²

The proposal arrives as pressure for AI guardrails grows, but it goes further than a conventional safety regime: rather than merely regulating superintelligence after it emerges, Sanders and Casar want to prevent it from being built at all.