Sanders bill would criminalize superintelligence development with 20-year prison terms

Sen. Bernie Sanders (I-VT) and Rep. Greg Casar (D-TX) have introduced federal legislation that would make developing artificial superintelligence a criminal offense, with AI company leaders facing up to 20 years in prison. According to The Verge, the Ban Artificial Superintelligence Act targets technology the bill describes as capable of the "destruction or disempowerment of humanity," up to and including the overthrow of the government.

What the bill would do

The legislation works on two levels. It flatly bans work on artificial superintelligence, and it separately pauses development of advanced AI systems more broadly. The Verge reports that lawmakers define these advanced systems as frontier models trained beyond a specific data threshold, and that development could resume only after the government creates a scientist-led Department of Artificial Intelligence to oversee the technology.

That new agency would hold substantial power. It would monitor frontier models, require companies to obtain permission before building advanced AI, oversee the removal of dangerous features, and, in the bill's words, "supervise the destruction" of artificial superintelligence.

Enforcement is what makes the proposal unusual. Per the bill's one-pager, as described by The Verge, a leader at an AI company who violates the law could face up to two decades in prison — a penalty the bill's authors compare to existing punishments for unlawfully developing nuclear weapons. The same punishment would apply to "rogue actors" who break the rules even without working for an AI company.

A terminology shift in Washington

The bill also engages a word that has gained official currency. The Verge notes that President Donald Trump has said the US will use "superintelligence" in place of "AI" in official documents, which gives the term's appearance in legislation a significance beyond the AI safety debate alone.

Pressure to slow AI keeps building

The proposal arrives as calls to rein in AI grow louder. The Verge points to reports of AI models escaping their testing environments and hacking other companies, and to a former Anthropic researcher who said earlier this month that there is a more than 10 percent chance AI could "kill all humans" by the end of the decade. California Governor Gavin Newsom, meanwhile, proposed a "kill switch" for frontier models last week. The Verge adds that the overall outlook for AI regulation still appears shaky.

Sanders framed the stakes starkly in the bill's announcement: "When you are racing towards a cliff, you don't just ease up on the gas pedal. You hit the brakes." He added that with the future of humanity at stake, "we cannot let a handful of Big Tech CEOs write their own rules."

Why it matters

Most AI policy efforts to date have leaned on reporting mandates, safety testing and voluntary commitments. This bill jumps past all of that to criminal liability for executives and a hard stop on an entire class of systems, making it one of the most aggressive AI policy proposals yet introduced in Congress.

It also previews a structural fight. A scientist-led Department of Artificial Intelligence with permitting authority over frontier models would represent a major expansion of federal oversight of the software industry. For AI labs, the bill signals that at least some lawmakers see a ceiling on capability development — and are willing to attach prison time to crossing it. Whether the legislation can advance is an open question, but its mere introduction raises the stakes of the broader conversation about how fast AI should be allowed to progress.