What the bill does
Senator Bernie Sanders and Representative Greg Casar introduced the Ban Artificial Superintelligence Act on 23 September 2026. Four provisions:
- Creates a cabinet-level Department of Artificial Intelligence. A new federal agency, at the same rank as Defense or Treasury.
- Bans the development of artificial superintelligence outright.
- Pauses advanced AI development until the new department establishes safety rules and review processes. No end date is specified.
- Sets criminal penalties of up to 20 years imprisonment for individuals, which the sponsors compare to unlawful nuclear weapons development, and a corporate death penalty for entities.
How it defines the thing it bans
The bill covers AI that "exceeds human cognitive performance and capabilities across most domains, or has sufficient capabilities to destroy or disempower humanity."
That is the definition in full. There is no FLOP threshold, no parameter count, no benchmark score, no capability evaluation standard. Every other serious AI regulation of the last two years - the EU AI Act, the export control regime, the US framework - has anchored to a measurable number precisely because "most domains" cannot be adjudicated.
A prosecutor would have to prove a model exceeds human performance across most domains, and a defendant would point at any domain where it does not. With 20 years of prison attached, that ambiguity is the central problem with the bill as drafted, and it is fixable - the sponsors could adopt the compute threshold language that already exists elsewhere.
What the sponsors said
Sanders: "When the future of humanity is at stake, we cannot let a handful of Big Tech CEOs write their own rules."
Casar: "Our bill bans the development of artificial superintelligence and pushes for international agreements so that no one, anywhere, builds AI too powerful for humans to control."
The international framing is the more practical half. A US-only ban moves frontier development rather than stopping it, and the bill's sponsors say so themselves.
What it means in practice
Very little in the near term. A bill introduced by a senator who does not chair the relevant committee, carrying a development pause and a new cabinet department, is a position statement rather than likely law. No co-sponsor list of consequence has been reported.
What it does mark is a shift in where the AI safety argument is being made. Through 2025 and most of 2026 the pause argument came from researchers and lab leaders. It is now in a bill, with prison terms in it, sponsored by elected officials - and it landed in the same month Meta publicly rejected industry-wide slowdown proposals and four labs faced cartel accusations over coordinated safety commitments.
FAQ
Would this ban ChatGPT or Claude?
Not as written. Current models do not meet the superintelligence definition. The development pause provision is the part that would affect existing labs, and its scope depends entirely on how "advanced AI development" gets interpreted.
What is the penalty?
Up to 20 years imprisonment for individuals and a corporate death penalty - dissolution - for entities.
Has anything like this passed before?
No. No jurisdiction has criminalised a category of AI development. The EU AI Act, the strictest regime in force, uses fines.
Who introduced it?
Senator Bernie Sanders and Representative Greg Casar, 23 September 2026.