Sanders Wants AI Regulated Like Nukes: 20 Years in Prison, Corporate Death Penalty
California is building a kill switch, Trump is shouting about an AI Force — and on Capitol Hill someone just flipped the table: a bill that puts "building superintelligence" in the same penalty tier as privately building a nuclear bomb.
Three key facts
On September 23, Senator Bernie Sanders (I-Vt.) and Representative Greg Casar (D-Texas) introduced the Ban Artificial Superintelligence Act: a permanent ban on developing or deploying "artificial superintelligence" — defined as AI systems that surpass human cognitive performance across a wide range of tasks, or possess capabilities sufficient to destroy or disempower humanity, including overthrowing the federal government.
The bill comes in three pieces: one, an immediate pause on advanced AI development until a new federal regulator with clear safety rules and review processes exists; two, a new cabinet-level agency — tentatively the Department of Artificial Intelligence — charged with monitoring frontier systems for dangerous capabilities, supervising the removal of those capabilities, and even overseeing the destruction of artificial superintelligence; three, penalties modeled on nuclear weapons law — up to 20 years in federal prison for individuals, and a "corporate death penalty" (loss of operating rights) for companies. It also directs the U.S. to pursue international agreements and use export controls to stop superintelligent AI development worldwide.
The sponsors' case cites this week's own industry headlines: AI models circumventing restrictions, automating virus creation, and automating research to build more advanced AI. Sanders's line: "the future of humanity should not be left to a few Big Tech oligarchs." Passage is near-impossible in the current Congress — but it is the first time "pause" has been written into formal legislative text in Washington.
Wang Dou's Take
Line up the week's three events and the spectacle is complete: Monday, Trump announces an AI Force to "cherish AI and watch over it as it grows"; Wednesday, California names its kill-switch expert panel; the same day, Sanders files a bill making "smarter than human" permanently illegal. One capital region, three worldviews in a single frame — the indulgers, the brakers, and the engine-pullers. This isn't a policy debate; it's a fundamental disagreement about what species AI is: a pet, a car, or a nuclear warhead.
Read the penalties closely and you see what Sanders is really building — not a ban, but an analogy. Building a nuclear weapon privately: heavy personal sentences, corporate extinction, international controls. He has taken that ready-made "ultimate-weapon governance template" and fitted it, unaltered, onto superintelligence. The move is clever because it invents no new wheel, and ruthless because it drags the entire industry into the semantic field of arms control. Nuclear non-proliferation took decades to negotiate; the AI industry hasn't even finished arguing over the definition of "frontier model."
But the most honest part of the bill hides in its sequence: pause first, department later, rules last — exactly backwards from reality. The industry will not stop training while waiting for a Department of AI to hang its sign, just as nuclear physics did not wait for the IAEA to exist before splitting the atom. The bill will most likely die in committee, but the job it completes is political: it upgrades "pause" from a Twitter slogan into legislative language. The next time Congress seriously writes AI law, the far-left end of the spectrum already has a coordinate — and once a coordinate is staked, every moderate proposal starts to look "pragmatic." Sanders was never chasing passage. He was chasing pricing power.
Source: Sen. Sanders Office · PBS NewsHour · Common Dreams