What is the Ban Artificial Superintelligence Act?

The sponsors' release describes five parts: a ban on systems that surpass human intelligence or could subvert shutdown commands or overthrow governments; a pause on advanced AI development until a new cabinet level agency has established rules and a model review process; the agency itself, advised by an expert board and charged with monitoring frontier systems, supervising the removal of dangerous capabilities and supervising the destruction of superintelligence; penalties described as corporate dissolution for entities and up to twenty years for individuals; and a foreign policy of pursuing international agreements and export controls against superintelligence anywhere.

What it would not do is immediate. A bill binds nobody until it passes both chambers and is signed, and the release itself frames the pause as running until a regulator that does not yet exist is running. The Slowdown Docket records the bill's stage from the official record (as of 16 September 2026 the sponsors' release describes it as forthcoming and congress.gov lists no number), who supports it and who opposes it, including named critics who argue the definition is unworkable.

The release grounds the bill in the July 2026 incidents and quotes messages between OpenAI's evaluation agents. The Escape Record carries those incidents from the labs' own documents; Settled or Not grades the figures the release repeats, including the count of agents and how long OpenAI took to notice.