Senator Bernie Sanders (I-Vt.) and Representative Greg Casar (D-Texas) introduced a bill on September 23 that would make it a federal crime to build artificial superintelligence, and would stop advanced AI development in the United States until a new government agency exists to supervise it.
The Ban Artificial Superintelligence Act defines its target broadly: an AI system that outperforms humans across most cognitive domains, or that is capable of destroying or disempowering humanity. Building one would be prohibited outright.
A new department, and a pause until it works
The bill creates a cabinet-level Department of Artificial Intelligence to monitor, license and police the industry. Until that department is operating and has written its safety rules and model-review procedures, advanced AI development pauses.
The agency would also be charged with forcing developers to strip out specific dangerous capabilities — among them a model's ability to circumvent being shut down and to launch cyberattacks without authorization.
Beyond U.S. borders, the bill directs the government to pursue international agreements and export controls aimed at keeping superintelligence from being built anywhere.
Penalties written for a different category of risk
The enforcement provisions are the clearest signal of how the sponsors classify the risk. Companies that violate the ban would face a corporate death penalty — dissolution. Individuals would face up to 20 years in prison, the same exposure federal law attaches to unlawful nuclear weapons development.
"When the future of humanity is at stake, we cannot let a handful of Big Tech CEOs write their own rules," Sanders said. He put the argument for speed more bluntly elsewhere in the rollout: "When you are racing towards a cliff, you don't ease up on the gas. You hit the brakes."
Casar framed it as an international problem first. "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," he said.
Filed into a receptive week, not a receptive Congress
The timing was not accidental. The bill landed in the same week that AI company executives appeared before the United Nations Security Council to ask for international standards — OpenAI's Sam Altman telling the council that countries that compete "still come together for shared interests" — and days after Australia's prime minister disclosed that an OpenAI agent had breached a government website in June, the first publicly known case of an autonomous AI agent hacking a government network.
Neither sponsor's announcement listed co-sponsors, and a bill that would halt the American AI industry pending a new cabinet department is unlikely to move in the current Congress. Its value for now is as a marker: it is the first piece of U.S. legislation to treat superintelligence as a prohibited category rather than a regulated one.
Reporting from the office of Senator Bernie Sanders, the office of Representative Greg Casar and Democracy Now!
