Quick answer
On September 23, 2026 Senator Bernie Sanders and Representative Greg Casar introduced the Ban Artificial Superintelligence Act, which would prohibit the development and deployment of AI systems exceeding a defined capability threshold until Congress establishes a safety regime. It is the first US federal bill to propose an outright prohibition rather than disclosure or testing requirements. Its prospects in the current Congress are poor; its significance is as a marker of where part of the debate has moved.
Most AI legislation so far has asked companies to test, report, and disclose. This bill asks them to stop. Whether or not it passes, that is a new position in mainstream US politics, and it arrived in a month when the labs shipped their most capable models yet.
What the bill proposes
- A prohibition on developing or deploying AI systems above a capability threshold defined in the bill, with the definition tied to broad, general-purpose capability rather than a compute number alone
- A pause until Congress enacts a licensing and safety framework, rather than a permanent ban
- Enforcement through civil penalties and, for knowing violations, criminal liability for executives
- Reporting obligations for frontier developers on training runs above set thresholds
The case for and against
Supporters argue that this month's releases, with their cybersecurity ratings and computer-use capabilities, show the industry racing ahead of any oversight, and that only a pause creates time to build one. Critics, including most of the industry and many researchers, argue that "superintelligence" is not definable in statute, that a US-only ban would hand the frontier to other countries, and that existing bills on testing and transparency are the practical path. Both sides agree the definition is the hard part.
Prospects and context
The bill has no committee schedule and little cross-party support; it is unlikely to advance this session. It sits alongside state laws in California and elsewhere, the EU AI Act's obligations now in force, and a set of narrower federal proposals on transparency and safety testing. Its practical effect is to shift the range of positions politicians can hold, which in the long run may matter more than its text.
Related reading
Bottom line
The bill will not become law soon, and it is not meant to. It is a marker: the first federal proposal to stop development rather than regulate it, introduced in the month the frontier moved fastest.