Sanders and Casar propose federal ban on AI superintelligence

Sen. Bernie Sanders (I-Vt.) and Rep. Greg Casar (D-Texas) formally introduced legislation Wednesday that would prohibit the development of artificial superintelligence and establish a new Cabinet-level agency to regulate AI systems.

The "Ban Artificial Superintelligence Act" targets what the lawmakers describe as the most dangerous capabilities in AI development, including recursive self-improvement—the ability of AI systems to autonomously build more advanced versions of themselves. The bill would also bar AI models from designing nuclear or biological weapons or attempting to evade human oversight.

Under the proposed legislation, violations would carry penalties comparable to illegal nuclear weapons development: up to 20 years in prison for individuals and what Sanders and Casar call a "corporate death penalty" for companies that circumvent federal oversight.

Why it matters

The proposal represents the most aggressive congressional attempt yet to regulate advanced AI, coming as lawmakers face mounting pressure to act following recent high-profile cyberattacks and warnings from industry insiders about existential risks. The bill's prospects remain uncertain in a divided Congress where Democratic leaders favor narrower regulatory approaches and some Republican leaders oppose AI safety legislation entirely.

New oversight structure

The legislation would create a "Department of Artificial Intelligence" led by a Senate-confirmed secretary. This new agency would be responsible for monitoring AI model capabilities, identifying risks, and pausing advanced development until adequate oversight mechanisms are in place.

Sanders framed the urgency in stark terms during a Wednesday press conference: "We cannot simply sit by idly and wait for advanced AI to cause the next 9/11 or Chernobyl or a global pandemic or a financial crisis before we take action to protect the American people and, of course, the people of the world."

Political headwinds

Casar expressed optimism that the bill "will be considered the commonsense proposal in the coming weeks" as public pressure mounts. However, the lawmakers face significant obstacles in both chambers.

The proposal comes as Senate Judiciary Chair Chuck Grassley (R-Iowa) separately pushes for passage of more modest AI whistleblower protections before the midterm elections. That bill, which would protect employees who disclose information about AI development risks, has bipartisan support but faces its own procedural hurdles.

The details were first reported by POLITICO.