Sanders and Casar Introduce Bill to Pause Advanced AI and Ban Superintelligence

A promised stop line for powerful AI is now a bill before Congress. Sen. Bernie Sanders and Rep. Greg Casar introduced the Ban Artificial Superintelligence Act on September 23. It would freeze advanced AI development until a federal regulator sets safety rules and permanently ban artificial superintelligence. Those restrictions are proposals, not rules in force, and the bill faces doubts about its prospects.

Sanders and Casar had announced the measure as forthcoming on September 3. They described the same temporary pause and permanent ban then. The September 23 introduction is the new step, not the first disclosure of their plan. It puts their proposed restrictions before lawmakers at a time when, according to SSBCrack News, some in Congress and industry doubt the bill will pass during this Congress.

A pause with an exit, and a ban without one

The proposed pause would stop development of advanced AI systems until a new federal regulator is operating and has established safety rules. That creates a path for work to resume under oversight. The superintelligence prohibition has no such endpoint: it would permanently bar development and use of that category of AI. The distinction is central to the bill, because the two restrictions would apply to different kinds of work for different lengths of time.

Where one category ends and the other begins would matter to developers. In their earlier description, the sponsors said the ban would cover systems that surpass human intelligence, could overthrow governments or could subvert shutdown commands. Those are the sponsors’ stated boundaries, not a settled assessment of how the bill’s language would be applied. AI policy expert Kevin Frazier, while welcoming congressional action, said the proposal leaves out independent bodies to verify AI safety.

The regulator would decide when work can restart

The bill calls for a Cabinet-level federal AI agency. Under the sponsors’ description, it would review the most capable systems for dangerous abilities, supervise the removal of those abilities and enforce the superintelligence ban. The agency would be advised by an expert board. It would also establish the rules and model-review process needed before the temporary development pause could end. Creating the department would therefore be part of putting the proposal into operation, not simply a change to Washington’s organization chart.

The proposed penalties give that agency’s enforcement role weight. A company that violates the measure could face forced dissolution; an individual could face up to 20 years in prison. The sponsors also want the United States to pursue international agreements to prevent superintelligence development elsewhere. Their earlier outline included coordination with allies and export controls among the possible approaches. Those are proposed tools for pursuing a worldwide ban, not agreements already reached.

A steep political test

Sanders argues that Washington is doing little to prepare for rapidly advancing AI. But Caleb Max, CEO of the National AI Association, called the bill unrealistic and warned that its penalties could face constitutional challenges. That is Max’s assessment, not a legal ruling. Sen. John Kennedy offered an even blunter forecast for AI legislation this Congress: “We’re not going to do anything on AI this Congress.”

For now, the introduction tests whether lawmakers will take up a bill built around stopping development before a regulator is ready. If they do, they will also have to grapple with the bill’s dividing line: which systems could resume development under safety rules, and which would remain prohibited.