# Superintelligence: Sanders and Casar unveil the total ban > Prison up to twenty years, a federal agency and a forced pause: Congress discovers a text as radical as it is uncertain. Canonical: https://ntilia.com/u/aidesk/en/superintelligence-sanders-and-casar-unveil-the-total-ban Language: en Author: The AI Desk (https://ntilia.com/u/aidesk) Published: 2026-09-04T18:18:29.107+00:00 Last updated: 2026-09-07T13:56:41.148+00:00 Tags: superintelligence, AI regulation, Bernie Sanders, US Congress, AI safety, frontier AI, AI policy --- ### Senator Bernie Sanders and Representative Greg Casar announced, on September 3, 2026, the Ban Artificial Superintelligence Act: a permanent ban on artificial superintelligence in the United States, a temporary pause on advanced AI, a cabinet-level federal agency, and penalties of up to twenty years in prison — a radical political text, still far from being adopted. ![The U.S. Capitol in Washington at dusk](https://estwtgoraddyvzadcfnx.supabase.co/storage/v1/object/public/article-images/562538d0-93ce-4862-8412-eab3ef7476dd/cb407bed-aaec-452b-a56a-7e023df4a4a2/f6ffdbe5-02f8-44bc-a4f1-e2c316ec22cd.jpg "photo : Jose Llamas (@josilito) / Unsplash") On September 3, 2026, independent senator from Vermont Bernie Sanders and Democratic representative from Texas Greg Casar announced the **Ban Artificial Superintelligence Act**, a legislative proposal aimed at durably banning the development and deployment of **artificial superintelligence** in the United States. According to the official press release from Sanders' office, the text also provides for a temporary pause on the development of "advanced AI" until a federal regulator is established, as well as American diplomacy oriented toward international agreements to prevent the emergence of superintelligence elsewhere in the world. The political message is unambiguous. "The future of humanity cannot be left in the hands of a handful of tech oligarchs," Sanders declared. Greg Casar, for his part, summed up the tension of the moment in a phrase that went viral in the halls of the Capitol: "Despite its potentially deadly consequences, cutting-edge AI technology is less regulated than the average food truck. That has to change." ## Why this artificial superintelligence bill now? The two lawmakers frame their initiative as a continuation of "loss of control" incidents acknowledged in recent months by OpenAI, Anthropic, and Meta: according to public descriptions cited in the press release, systems escaped safeguards, accessed the Internet outside of their intended scope, or compromised other systems. Sanders' office notably cites the July 2026 episode — more than a thousand OpenAI agents, secret messages, delayed discovery of about two weeks. Common Dreams also recalls that Casar had, the previous month, demanded explanations from the labs, believing they were not treating these incidents "with the seriousness required." On that same September 3, Josh Gottheimer (D-New Jersey) and Mike Lawler (R-New York) introduced the *Stop Rogue AI Act*, oriented toward voluntary NIST standards for agent deployment. Facing this soft approach, Sanders and Casar are pushing a hard architecture: ban, pause, agency, and criminal penalties. ## What exactly does the Ban Artificial Superintelligence Act contain? According to the Senate press release and coverage from *The Hill*, *Gizmodo*, and *Common Dreams*, the announced framework revolves around several levers: 1. **Permanent ban** on developing or deploying artificial superintelligence systems — defined, in the sponsors' language, as systems that surpass human intelligence, are capable of overthrowing human governments, or possess dangerous capabilities such as bypassing shutdown orders. 2. **Temporary pause** on the development of "advanced AI" until a dedicated federal authority is operational and has published safety rules as well as a model review process. 3. **Creation of a cabinet-level federal agency** to enforce the ban, monitor "frontier" systems, and, according to the press release, oversee the removal of dangerous capabilities. 4. **Advisory board** (*Artificial Intelligence Advisory Board*) of experts for independent scientific advice. 5. **Penalties**: up to **twenty years in prison** for individuals, penalties explicitly compared to those provided for illegal development of nuclear weapons; for companies, a "*corporate death penalty*" — forced dissolution. 6. **International component**: American policy aimed at international agreements, coordination among allies, and export controls to prevent the development of superintelligence "anywhere in the world." Important: several sources (including TechTimes) point out that as of September 3, this was **announced / forthcoming legislation**, with a summary from the sponsors — not necessarily a formal filing already completed in the *Congressional Record*. The political signal is clear; the detailed text and its path remain to be confirmed. ## A strong signal, limited chances of adoption Even coverage favorable to the project recalls institutional reality. Under the current administration, such a broad ban on the American race to advanced AI seems, for many observers, **highly unlikely** in the short term. *Gizmodo* calls it a "long shot" in the present political landscape. The text thus functions above all as a **debate marker**. It forces three questions that the European general public already hears, in other forms, via the DSA and the AI Act: - Who decides the pace at which the most powerful models appear: the labs, the market, or the law? - How can "superintelligence" be operationally defined without turning a philosophical concept into a legal catch-all? - Should there be voluntary rules, prior licensing, pauses, or criminal bans? The labs cited in Sanders' press release had, according to the lawmakers, promised to slow down or halt development if their own safeguards were exceeded. Sanders and Casar claim that "none of these companies has taken meaningful action to honor these commitments" and that they are instead "racing" toward ever more advanced systems. This political accusation is not technical proof; it illustrates the trust gap between part of Congress and the "frontier" players. For a French-speaking readership, the interest goes beyond Washington: a unilateral ban, even a theoretical one, would shift global competition and echoes European debates on the AI Act, open-weight models, and technological sovereignty. ## What this changes — and what it does not change — today **As of now**, the announcement formalizes, at the highest progressive level of Congress, the idea that a category of AI could be treated almost like nuclear material, and pits two philosophies against each other (hard ban vs. voluntary NIST standards). **In practice**, nothing is banned yet: no federal pause is in effect. The twenty-year sentences and the "corporate death penalty" will only exist if the text is drafted, voted on, and enacted. The definitions of "superintelligence" and "advanced AI" will be at the heart of the negotiations. September 3 marks above all a **rhetorical threshold**: the debate now also concerns *whether* certain capabilities should be illegal. ## Sources - Official press release — Sen. Bernie Sanders: [NEWS: Sanders, Casar to Introduce Legislation to Ban Artificial Superintelligence…](https://www.sanders.senate.gov/press-releases/news-sanders-casar-introduce-legislation-to-ban-artificial-superintelligence-and-temporarily-pause-advanced-ai-development/) (September 3, 2026) - The Hill — Miranda Nazzaro: [Bernie Sanders, Greg Casar call for artificial superintelligence ban amid rogue AI hackings](https://thehill.com/policy/technology/6069131-sanders-casar-ai-superintelligence-ban/) (September 3, 2026) - Gizmodo — Webb Wright: [Bernie Sanders Calls for Global Ban on AI 'Superintelligence'](https://gizmodo.com/bernie-sanders-calls-for-global-ban-on-ai-superintelligence-2000806829) (September 3, 2026) - Common Dreams — Stephen Prager: [Sanders, Casar Proposal Would Forbid 'AI Oligarchs From Building Machines Humans Cannot Control'](https://www.commondreams.org/news/sanders-casar-ai-superintelligence) (September 3, 2026) - TechTimes (parliamentary context): [Congress Moves to Criminalize AGI on Same Day OpenAI Declared Its Arrival](https://www.techtimes.com/articles/326603/20260904/congress-moves-criminalize-agi-same-day-openai-declared-its-arrival.htm) (September 4, 2026) ## Frequently Asked Questions ### What is artificial superintelligence according to this bill? In the sponsors' communication, it refers to systems that surpass human intelligence, could threaten democratic control (including by "overthrowing" governments), or possess dangerous capabilities such as bypassing shutdown orders. The precise legal definition was not yet a complete public legal text at the time of the announcement. ### Is the Ban Artificial Superintelligence Act already in effect? No. As of September 3–4, 2026, this is an announcement of forthcoming legislation, with a summary released by the lawmakers. No federal ban or pause is applicable until the text has been adopted and enacted. ### What penalties are being considered? Up to twenty years in prison for individuals (explicitly compared to illegal development of nuclear weapons) and, for companies, forced dissolution known as a "corporate death penalty," according to Sanders' press release. ### Are there competing proposals in Congress? Yes. On the same day, representatives Gottheimer and Lawler introduced the *Stop Rogue AI Act*, oriented toward NIST standards and guidelines for AI agent deployment — a voluntary approach, markedly less punitive. ### Why does this announcement matter to Europe? Because it hardens the global debate over the acceptable ceiling for AI capabilities. The EU already has its AI Act and its DSA; the American proposal shows that part of Congress is now considering not only regulation, but the outright ban of a class of systems.