Bernie Sanders proposes banning ‘superintelligence’ and putting violators in prison

Sen. Bernie Sanders and Rep. Greg Casar have unveiled legislation that would prohibit the development of artificial superintelligence and halt work on high-end AI systems until new federal oversight is in place. The bill describes superintelligence as technology capable of destroying or disempowering humanity, including by toppling governments, and seeks to make development of such systems a felony.

Under the proposal, the government would create a Department of Artificial Intelligence, led by scientists, to supervise advanced AI research. Companies would need explicit permission from that agency to continue building models considered ‘‘frontier’’ or advanced — defined in the bill by a specific training-data threshold — and the department would be empowered to remove hazardous capabilities and oversee destruction of illicit systems.

The legislation assigns criminal liability to corporate leaders who proceed without authorization, exposing them to up to 20 years behind bars; the bill’s summary notes this is comparable to penalties for unlawfully developing nuclear weapons. The penalty would also extend to unaffiliated “rogue actors” who carry out prohibited development outside established organizations.

The measure arrives amid intensifying debate over AI safety and governance. The Verge notes related developments including President Trump’s stated preference to use the term “superintelligence” in official documents, public calls for slower development of frontier models, a proposal from California’s governor for a ‘‘kill switch’’ on powerful systems, and recent comments from a former Anthropic researcher expressing severe risk estimates for advanced AI.