US bill would pause frontier AI until a federal department approves it

Senator Bernie Sanders and Representative Greg Casar have introduced a bill that would prohibit artificial superintelligence and pause development of advanced AI systems until a new federal regulator is operating. The proposal defines the prohibited technology in terms of humanity’s destruction or disempowerment, including the ability to overthrow government.

The Ban Artificial Superintelligence Act would create a scientist-led Department of Artificial Intelligence. Developers would need its permission to train advanced systems above a specified data threshold. The department would monitor frontier models, require dangerous capabilities to be removed and supervise the destruction of systems classified as artificial superintelligence.

Executives who knowingly continue prohibited development could face up to 20 years in prison, a penalty the proposal compares with existing punishment for unlawfully developing nuclear weapons. The same exposure would apply to independent actors outside a company. The article does not report committee action or support sufficient to pass the measure, so it remains a legislative proposal rather than an imminent rule.

The bill arrives amid competing approaches to AI oversight, including California’s proposed frontier-model kill switch. It represents one of the most restrictive options: development would stop first, then resume only under a newly created federal approval system.