Discussion draft: a bill to ban superintelligent AI and regulate its precursors.

Major AI companies are attempting to develop artificial superintelligence (ASI). Superintelligent AI systems would be able to fully replace and outmatch human individuals and organizations across all tasks, including in domains like intelligence, cyber operations, technological research and development, and military strategy.

Superintelligent AI systems would operate fully autonomously, and AI companies admit there is no known way to keep them under control should they be developed. Superintelligence would therefore not be a strategic weapon like nuclear weapons, but rather a threat actor in and of itself, capable of threatening the national security of the United States. In fact, Nobel Prize winners, leading AI scientists, and even the CEOs of leading AI companies have repeatedly warned that superintelligent AI systems pose a risk of human extinction.

This draft bill proposes a framework to decisively eliminate this threat to America’s national security by criminalizing the development of superintelligence domestically and extraterritorially, and regulating the precursors that could be used to build it.

This framework targets the security risks of superintelligence head on while aiming to preserve American leadership in beneficial AI. Ultimately, in addition to enacting legislation like this framework, America should lead efforts to establish an international ’trust but verify’ regime backed by credible deterrence to prohibit superintelligent AI development around the world.