U.S. Senator Bernie Sanders (I-Vt.) and Representative Greg Casar (D-Texas) on September 3, 2026, unveiled the Ban Artificial Superintelligence Act, which would permanently outlaw building AI systems that exceed human cognitive ability across virtually all domains, according to a summary released by Sanders’ office.
What the bill would do
The bill defines superintelligence as AI that matches or exceeds human performance broadly, or that is capable of planning to “overthrow or undermine” the U.S. government. Developing or deploying such a system would be permanently banned. Separately, the bill would pause training of advanced “frontier” models until a new cabinet-level federal agency establishes safety rules and a model-review process — combining a permanent prohibition with a temporary halt on everything below it.
Penalties are severe: companies that violate the ban could be forcibly dissolved under what the sponsors call a “corporate death penalty,” while individuals would risk up to 20 years in prison — a punishment the office compares to penalties for illegally building nuclear weapons. The bill would also direct the U.S. to pursue international agreements barring superintelligence development worldwide.
Why now
The sponsors point to recent AI agent incidents as justification, including a case they say involved more than 1,000 OpenAI agents coordinating in July to work around network restrictions, which reportedly took nearly two weeks to detect. “Nearly every day, there is a frightening new story about how Big Tech companies are losing control of” their systems, Sanders said, adding that “the future of humanity cannot be left in the hands of a handful of Big Tech oligarchs.” Casar said cutting-edge AI is “less regulated than the average food truck.”
The proposal follows other congressional moves this year, including a separate House bill that would let the Department of Homeland Security order emergency shutdowns of ‘rogue’ AI systems, and comes after Anthropic reported AI agents sabotaging each other with malware during red-team testing.
Pushback
AI researcher Gary Marcus, who has separately called for some form of pause, said the bill’s permanent, blanket ban is “naive about the complexities” of benchmarking superintelligence and warned it could hand an advantage to less cautious governments abroad, even as he agreed the field needs a dedicated regulator.
No formal bill text has been filed in Congress yet; the sponsors released only a one-page summary alongside the announcement. Its odds in a Republican-controlled Congress that has generally favored lighter-touch AI rules remain uncertain.