FULL STORY

The Ban ASI Bill: From Threat to Formal Proposal

Senator Sanders announced plans to ban artificial superintelligence, with bill details emerging over weeks—including up to 20 years in prison and a training pause—followed by a parallel UK bill and public reactions from figures like Jensen Huang.

2026-09-04 ~ 2026-09-27 · 11 episodes · 116 posts

Episode 1 · Sanders Bill Would Permanently Ban Superintelligence, With Up to 20 Years in Prison (2026-09-04, 21 posts)

On September 4, Senator Bernie Sanders and Rep. Greg Casar introduced the Ban Artificial Superintelligence Act, which would permanently ban the development or deployment of AI systems at or beyond human cognitive ability. Individuals could face up to 20 years in prison, with penalties modeled on laws against illegal nuclear weapons development, and companies face corresponding severe penalties. The bill would also halt advanced AI development until a new government AI oversight body is established, and push international agreements to stop other countries from building superintelligence. Per Axios, it is one of the strongest statements on frontier AI risk from US politics to date, and it has sparked unusually intense division in the AI safety and tech communities.

Confirmed

  • The bill, with legislative text, permanently bans superhuman AI; individuals face up to 20 years, penalties reference nuclear weapons law, and companies face heavy penalties
  • It requires a full pause on advanced AI development before a new government AI oversight agency exists, and seeks international agreements to block ASI development abroad
  • The bill justifies itself by claiming ASI threatens Americans' safety, freedom and lives, and that AI industry leaders themselves admit building uncontrollable dangerous technology
  • Longtime AI critic Gary Marcus publicly opposed the bill: he supports an AI regulatory agency (he testified before the Senate in 2023) and could support a pause, but a permanent, unilateral ban on all superhuman AI research is too broad and wrong-headed
  • SingularityNET founder Ben Goertzel wrote that the bill is "half right on diagnosis, almost entirely wrong on remedy": he agrees humanity's future shouldn't be decided by tech oligarchs, but a ban won't stop AGI, only push its development elsewhere
  • Developer csuwildcat strongly objected, arguing AI is increasingly capable of novel medical treatments and banning it condemns patients; poster Kye Gomez blasted the bill, and one sharer sarcastically suggested "banning senators dumber than computers"

Unconfirmed

  • A poll cited by Cambridge researcher David Krueger claiming most Americans support pausing AI remains unpublished; details unknown
  • The bill's legislative prospects and implementation details remain unclear

Why it matters

  • This is the harshest legislative attempt targeting ASI yet. Reddit user the320x200 noted it shows "I can download open models so legislation doesn't matter" thinking no longer holds—lawmakers intend to keep powerful models out of reach
  • MIRI chief Nate Soares remarked it was "unthinkable a year ago," reflecting rapid mainstreaming of AI safety; Marcus's "pause yes, ban no" and Goertzel's stance show sharp disagreement over ban scope even among AI critics
  • Industry voices (relayed via YC partner Aaron Bergman) argue unilateral US legislation may have limited effect without global coordination; supporters counter with concerns that banning strong AI forfeits medical and other application benefits

1 more related posts →

Episode 2 · Sanders: Rogue Superintelligence Is All Humanity's Problem (2026-09-04, 3 posts)

Senator Bernie Sanders warned that a rogue superintelligent AI would be a problem for all humanity, not just the US or China, and called for global cooperation to prevent such a scenario.

Episode 3 · UK Parliament Sees World's First Anti-Superintelligence Bill as US Follows Suit (2026-09-08, 8 posts)

Between September 8 and 10, parallel legislative moves targeting superintelligence emerged in the UK and the US. UK MP Alex Sobel formally introduced the Artificial Superintelligence Security Bill, drafted by advocacy group ControlAI, with cross-party support; per the proposers and multiple accounts, it is the world's first bill introduced in a legislature aimed at banning superintelligence development. Almost simultaneously, US Senator Bernie Sanders announced the Ban Artificial Superintelligence Act, with Rep. Greg Casar involved. ControlAI claims three legislative breakthroughs in two weeks across both countries, with a global treaty also being pursued. All bills remain at the introduction stage.

Confirmed

  • The bill was introduced in the UK Parliament by MP Alex Sobel, drafted by ControlAI, with cross-party support
  • It seeks to ban superintelligence development while ensuring the UK continues to pursue AI
  • The proposers cited risks such as rogue AI clusters escaping, launching cyberattacks, and hacking multibillion-dollar companies
  • ControlAI says it has held over 350 meetings with legislators
  • Per accounts citing TIME, Senator Sanders announced the Ban Artificial Superintelligence Act to bar AI systems smarter than humans and pause other advanced AI research until new safety rules exist, with Rep. Greg Casar involved; the UK bill is said to be the first of its kind in a G7 parliament
  • ControlAI claims three legislative breakthroughs in two weeks and a push toward a global treaty

Unconfirmed

  • The full text and enforcement mechanisms of the US Ban Artificial Superintelligence Act
  • The specific parties and timeline for the proposed global treaty

Why it matters

  • If advanced, these would be among the world's first legislative bans on superintelligence development, setting a precedent for other countries' AI safety laws
  • The bills attempt to draw a line between banning superintelligence and continuing AI development—an innovative legislative approach
  • Synchronized US-UK action signals a shift from single-country advocacy toward cross-national legislative coordination
  • Multiple reposts (@DavidSKrueger, @zetalyrae, @austinc3301, @ChrisArmstrong) cite the same sources, with information concentrated on September 8–10

Episode 4 · Sanders Moves to Ban Superintelligence, Sparking AI Debate (2026-09-09, 3 posts)

US Senator Bernie Sanders announced plans to legislate a ban on superintelligence and pause AI development, citing developers' own warnings. AI safety advocates back the move, while critics like Matthew Berman call regulation the wrong answer.

Episode 5 · 71 UK Lawmakers Back Bill to Ban Superintelligent AI (2026-09-12, 10 posts)

On September 12, Labour MP Alex Sobel announced that 71 cross-party MPs and peers had signed in support of his Artificial Superintelligence Security Bill, urging the UK government to ban the development of superintelligent AI (ASI) and broker an international agreement. Signatories span Labour, Conservative, Lib Dem and SNP, including 15 former ministers and former Cabinet Secretary Robin Butler. Campaign group ControlAI published the bill's details, calling it the world's first ASI ban formally tabled in a legislature and proposing a G20-driven international ban. The initiative explicitly preserves the UK's strategic AI ambitions rather than opposing AI wholesale.

Confirmed

  • 71 cross-party parliamentarians signed the call to ban ASI development and urged the Prime Minister to adopt the bill
  • Signatories include 15 former ministers and Robin Butler across Labour, Conservative, Lib Dem and SNP
  • ControlAI's published details: the first such bill in a legislature, directly banning ASI R&D and pursuing a G20-coordinated global ban, citing recent AI-control incidents
  • Sobel frames the move as guarding against extinction risk while keeping strategic AI ambitions

Unconfirmed

  • Several posts name addressee Andy Burnham as Prime Minister, inconsistent with public records and pending verification; one post cites 40 signatories versus Sobel's 71
  • No follow-up yet on the bill's parliamentary prospects or G20 traction

Why it matters

  • If passed, it would directly constrain frontier AI R&D in UK labs; @Patrice argues UK labs are in serious trouble
  • A rare collective legislative call to ban ASI outright, with potential G20 spillover as a template for other countries
  • James Wise calls the position self-contradictory and damaging to UK competitiveness; Gary Marcus argues the ban should target poorly aligned AI, not superintelligence itself
  • RobertHaisfield dismisses the move, noting only two countries' stances truly matter on pausing/banning ASI, and the UK is not one of them

Episode 6 · Viral Claim: Sanders AI Bill Would Jail Frontier AI Developers Up to 20 Years (2026-09-12, 12 posts)

On Sept 12, a widely shared post claimed that a draft AI bill by U.S. Sen. Bernie Sanders would impose up to 20 years in prison on developers and researchers working on 'advanced AI,' triggering strong backlash across Silicon Valley and the AI community. As of the available material, all reporting is secondhand retweeting with no primary legislative text, so the provision's authenticity remains unverified.

Confirmed

  • The circulating claim: researchers who violate or circumvent restrictions on 'advanced AI' development could face up to 20 years in prison—far harsher than any prior mainstream AI regulatory proposal
  • The claim originated with @venturetwins and spread widely: @xiaosun86 called Sanders a 'lunatic' and said he was glad Sanders lost the primary; @HankYeomans said he had missed this detail and questioned its seriousness; retweeters mocked it with 'commie gulag' memes as innovation-blocking overregulation
  • In content relayed by @ivanbezdomny, industry figures including Suchenzang attacked the draft as a product of safetyists' envy toward capability-research engineers
  • @Hesamation compared the 20-year sentence to penalties for bank robbery or supporting terrorism, expressing shock; @ShikharMurty shared it jokingly
  • @tzmartin noted the draft's potential impact on the open-source ecosystem, with commenters calling the provision overly aggressive
  • Separately, a Reddit screenshot relayed by @Cagnazzo82 described a 'Stop AI' bill, backed by politicians and anti-AI advocates, that would criminalize developing superintelligence with up to 20 years in prison—called one of the most radical superintelligence-regulation proposals

Unconfirmed

  • The actual text, details, and legislative status of the Sanders draft; no primary legislative document has surfaced
  • Whether the 'Sanders draft' and the 'Stop AI' bill are the same or related legislation cannot be determined from the material

Why it matters

  • If true and advanced, such provisions would reshape the legal risk structure of frontier AI research, hitting open-source communities especially hard
  • It signals that radical U.S. superintelligence-regulation proposals are entering public discourse, drawing fierce pushback from researchers and practitioners

Episode 7 · Sanders Proposes Bill to Ban Superintelligence and Pause Advanced AI (2026-09-12, 2 posts)

US Senator Bernie Sanders and Representative Greg Casar introduced the Ban Artificial Superintelligence Act, calling for a full pause on advanced AI development and a legislative ban on superintelligence, marking one of the strongest political stances against rapid AI progress.

Episode 8 · Sanders Proposes 20-Year Prison Terms for ASI Developers (2026-09-14, 2 posts)

Senator Bernie Sanders has proposed sentencing AI developers working toward artificial superintelligence to up to 20 years in prison, comparable to penalties for illicit nuclear weapons development.

Episode 9 · Jensen Huang on Ezra Klein Show: Shut down labs that can't control their models (2026-09-23, 27 posts)

NVIDIA CEO Jensen Huang appeared on The New York Times' Ezra Klein Show, making forceful statements on AI safety and clashing repeatedly with the host. The transcript was published by NYT and NVIDIA's official account reshared the full episode, sparking wide debate in the AI safety community.

Confirmed

  • Huang argued that fear of AI is "getting way out of hand" and that he does not want people to fear the technology; he said AI poses no existential threat regardless of how many former lab employees warn otherwise (per @ShakeelHashim, @firstadopter, @AlexTensor).
  • He directly attacked OpenAI, Anthropic and AI doomers, saying no one builds more compute than those calling to slow down, and rebutted Geoffrey Hinton's 10% human-extinction estimate. He also dismissed Hinton's prediction that AI would replace radiologists as irresponsible and lacking scientific basis, noting Hinton's past predictions have not come true (per @Hesamation, @AlexTensor, @rohanpaulai).
  • Addressing labs' loss-of-control claims, Huang said that if Anthropic, OpenAI, Google or xAI believe they are out of control, they should stop releasing products until they are in control. When Klein described an Anthropic data-leak incident, Huang responded that the product shouldn't have been released; after Klein noted it was never released, Huang went further: labs that admit they cannot control their experiments should be shut down (per @haider, @ShakeelHashim, @dhadfieldmenell, @notkilleveryoneist, @noamyy). Ethan Perez's shared clip highlights Klein's reply that labs withheld the model precisely for that reason, exposing the tension between safety-motivated delays and the chipmaker's stance.
  • The "shut down the labs" remark spread widely via Midas Project and others as a sharp rebuke of AI-safety rhetoric; Sneha Revanur noted Huang prefers shutting labs over new legislation, contrasting with figures like YC's Garry Tan (per @ShakeelHashim, @austinc3301).
  • NVIDIA's official account shared the full interview with its position: AI is a technology built and improved by people, developers are responsible for deploying it carefully, and helping people benefit from AI means taking its challenges seriously (per @nvidia).

Unconfirmed

  • Former OpenAI policy lead Miles Brundage reshared the remarks, arguing those reading them as Huang conceding or being outmaneuvered miss the substance, but his full argument is not detailed in the material (per @MilesBrundage).

Why it matters

  • As head of the world's largest AI compute supplier, Huang's public stance on AI safety and doomerism is a bellwether, and his rare named criticism will sharpen the accelerationist-versus-safety divide.
  • Critics note the irony that he calls for shutting down labs that can't control their experiments while NVIDIA invests tens of billions of dollars expanding compute; @haider mocked it as a CEO simultaneously hyping loss-of-control risk while asking to relax antitrust and liability laws.

7 more related posts →

Episode 10 · Sanders Bill Proposes Permanent Ban on Superintelligence, Up to 20 Years in Prison (2026-09-23, 23 posts)

On September 24, Senator Bernie Sanders and Rep. Greg Casar introduced the Ban Artificial Superintelligence Act, proposing a permanent ban on superintelligence, a pause on frontier AI development, up to 20 years imprisonment for violators, and asset forfeiture for companies. The bill remains a proposal with uncertain legislative prospects, but its radical provisions triggered polarized reactions across the AI community, making it one of the most aggressive congressional attempts to restrict frontier AI.

Confirmed

  • Per Krueger's endorsement thread and rohanpaulai's summary, the bill's core provisions include: creating a federal Department of AI; permanently banning "superintelligence" (systems exceeding human cognition in most domains or capable of severe harm); banning "superintelligence precursors" including automated AI research and recursive self-improvement; pausing training and deployment of "advanced AI" models at or above 10^25 FLOP/operations until the department establishes governance; asset forfeiture and up to 20 years imprisonment; and pushing international coordination.
  • Cambridge AI safety researcher David Krueger publicly endorsed the bill, declaring an "AI emergency"; MIRI also endorsed it, as did employees at several leading AI companies. PauseAI launched a phone-calling campaign urging constituents to press their representatives.
  • NBC News covered the proposal. Polymarket prices the odds of a US AI safety bill passing before 2027 at only 13%, with about $165K in volume.

Unconfirmed

  • An analysis by @1a3orn of the draft PDF suggests the ban covers not only superintelligence but any system displaying "superintelligence precursor characteristics," potentially capturing many current models—this reading awaits legislative testing.

Why it matters

  • If the precursor-characteristics clause were enacted, it could directly restrict mainstream model development and deployment; the ban on recursive self-improvement targets a current mainstream research direction.
  • Reactions split sharply: Krueger and MIRI urged emergency action; DavidLinthicum called it bipartisan "fear-driven legislation"; VraserX argued a permanent ban is a terrible way to share AI-created wealth, questioning why protecting workers means preserving the necessity of work; David Sacks' camp criticized overregulation; prediction markets price low passage odds.
  • Margaret Mitchell pressed Krueger on the distinction between "AI escaping our control" and "we developing AI in an out-of-control way," arguing this subtle divergence is fracturing the safety community's legislative narrative.

3 more related posts →

Episode 11 · Sanders ASI Ban Bill Draws Fire from Safety Researchers and Sacks (2026-09-26, 5 posts)

Sen. Sanders and Rep. Casar introduced the 2026 Ban Artificial Superintelligence Act, which would create a federal Department of AI (DoAI) and pause training of models above 10^25 ops until the department is ready. Harvard researcher Stephen Casper posted a long thread analyzing the bill section by section, concluding that its boldness may be one of humanity's best chances to avert catastrophic AI risk, though its design is flawed with vague, hard-to-enforce definitions.

Confirmed

  • The bill would establish a federal DoAI and pause training of models exceeding 10^25 ops until it is operational.
  • Models above 10^25 ops face licensing-related requirements, per Casper's breakdown.
  • Former White House AI lead David Sacks criticized the bill's overly broad definition of "superintelligence," arguing today's models might already qualify and open-source models would be caught; he also revealed the bill would impose up to 20 years in prison for developers.

Why it matters

  • The bill is among the most aggressive attempts in Congress to regulate frontier AI, and if advanced it would directly affect large-model training and the open-source ecosystem.
  • Casper and Sacks, from AI-safety and industry/policy perspectives respectively, offer sharply divergent assessments, exposing the core tension between decisive risk prevention and overbroad definitions.
  • Casper stated he voices his true views even outside the Overton window, indicating his support for decisive regulation is not performative positioning.