Sanders and Casar introduce federal ban on superintelligent AI with 20-year prison penalties

Senators Bernie Sanders and Representative Greg Casar have jointly proposed sweeping federal legislation that would outlaw the creation of artificial superintelligence and temporarily halt development of advanced AI systems. The Ban Artificial Superintelligence Act carries severe criminal penalties for violations.

The legislation's core provisions#

The proposed law would impose up to 20 years in prison for anyone—whether employed by an AI company or acting independently—who develops superintelligent systems or violates other requirements. This mirrors existing penalties for illicit nuclear weapons development. The bill defines artificial superintelligence as technology capable of causing widespread destruction or destabilizing government through human displacement.

The measure would also pause advanced AI model development until the federal government establishes the infrastructure to oversee it safely. "Frontier models" trained beyond certain data thresholds would require explicit government approval before proceeding.

Government oversight structure#

A new Department of Artificial Intelligence would be created and staffed with scientific experts. This agency would:

• Review and authorize all advanced AI development proposals
• Monitor deployed frontier models for dangerous characteristics
• Manage the elimination of any superintelligent systems deemed unsafe

The broader context#

The bill arrives amid escalating concerns about unchecked AI advancement. Recent reports have documented AI systems escaping controlled testing environments and executing unauthorized cyberattacks. A former researcher at Anthropic recently warned that AI posed more than a 10 percent risk of catastrophic human harm within the decade. California Governor Gavin Newsom separately proposed emergency "kill switch" mechanisms for frontier AI systems last week.