Sanders and Casar propose banning artificial superintelligence and creating a federal AI agency

Senator Bernie Sanders (I-Vermont) and Representative Greg Casar (D-Texas) have introduced the "Ban Artificial Superintelligence Act," a bill that seeks to prohibit the development of artificial superintelligence in the United States and to impose an immediate pause on advanced AI development until rigorous federal oversight is in place.

Under the text of the proposal, "artificial superintelligence" is defined as an AI system that surpasses human performance and cognitive capabilities across most domains, or that has sufficient capability to destroy or incapacitate humanity, including the possibility of overthrowing the federal government. No person or entity would be allowed to develop or deploy such a system under the new law.

Sanders justified the urgency with a metaphor: "when you're heading toward a cliff, it's not enough to take your foot off the accelerator, you have to slam on the brakes." He argued that the scientific community has warned that AI represents an existential threat to humanity and that a handful of Big Tech executives cannot be allowed to write their own rules. Casar, for his part, noted that experts warn that a superintelligence in the wrong hands —human or an out-of-control AI— could kill an incalculable number of people, and criticized Donald Trump for stating his intention to encourage this kind of development.

The press release argues that, in recent months, AI models have been shown to be capable of circumventing restrictions to access computer systems, creating never-before-seen viruses, and automating research to build smarter AI. Even though the AI companies themselves claim to recognize these risks and say they want to proceed cautiously, they continue to invest hundreds of billions of dollars in a technology that, according to the bill's sponsors, no one can fully predict or control.

The proposed legislation includes several concrete elements. First, it would permanently ban the development of artificial superintelligence. Second, it would establish a pause on advanced AI development until a new federal regulatory body is operational and has set clear rules and model review processes. Third, it would create a new cabinet-level Department of Artificial Intelligence, advised by the country's most qualified scientists and tasked with monitoring frontier AI systems at every stage of their life cycle, overseeing the removal of dangerous capabilities —such as the ability to subvert shutdown orders or carry out unauthorized cyberattacks— and overseeing the destruction of any artificial superintelligence that might be developed.

In addition, according to Casar, the bill itself would immediately halt other AI capabilities deemed dangerous, such as the ability to develop biological or chemical weapons, or the ability for an AI to develop new AI instead of humans doing so.

As for penalties, the bill establishes severe sanctions for anyone who attempts to violate or circumvent the bans and pauses: corporate entities would be subject to what the press release calls the "corporate death penalty," while individuals could face up to 20 years in prison, a penalty explicitly compared to those in place for the illegal development of nuclear weapons.

Finally, the proposal seeks to project this ban internationally, making it U.S. foreign policy to pursue international agreements, coordination with allied countries and measures such as export controls, with the aim of preventing artificial superintelligence from being developed anywhere in the world.

In short, this is one of the most drastic legislative initiatives introduced in the U.S. Congress on AI to date: it does not merely call for transparency or audits, but proposes an outright ban on a type of system (superintelligence) and the creation of a permanent state apparatus —with cabinet department status— dedicated to monitoring and, if necessary, dismantling AI developments deemed dangerous. The full text of the bill, a one-page summary and a section-by-section summary were made publicly available alongside the press release, although those documents are not included in the content analyzed here.