Ban it, build it, or build it carefully?

The loudest call for a ban is a short public statement run by the Future of Life Institute, a non-profit that campaigns on AI risk. Launched in October 2025, it reads: "We call for a prohibition on the development of superintelligence, not lifted before there is broad scientific consensus that it will be done safely and controllably, and strong public buy-in." Anyone can sign. On 17 October 2026 its page showed 77,670 signatures, including a petition's 5,000. Counts from other sources differ, so cite the live page and its date.

Signers named in press coverage include the AI researchers Yoshua Bengio and Geoffrey Hinton, and also public figures from across politics, which is part of the point: the campaign wants to show the concern is not confined to one camp. One of its leading supporters, the computer scientist Stuart Russell, has said it is not a ban in the usual sense, since it is conditional on safety and public support.

The bills

• United States. According to Al Jazeera and the GovTrack tracker, Senator Bernie Sanders introduced the Ban Artificial Superintelligence Act of 2026 (S. 5493) on 23 September 2026, with a companion bill from Representative Casar. Reports say it would prohibit development and create a new federal agency. It was referred to the Senate Commerce Committee.
• United Kingdom. A private member's bill to prohibit artificial superintelligence was introduced in the House of Commons and had its first reading on 8 September 2026, according to news and law-firm summaries. The government said it did not consider the proposed measures to be the right approach. Private members' bills very rarely become law.

Red lines

A separate civil-society effort, the Global Call for AI Red Lines, was launched at the UN General Assembly in September 2025 by three organizations. It asks governments to agree international red lines by the end of 2026. It is a campaign, not a government instrument.

• Meta wrote in July 2025: "Meta's vision is to bring personal superintelligence to everyone" and "Developing superintelligence is now in sight". It added that superintelligence "will raise novel safety concerns" and that it would be "careful about what we choose to open source".
• Safe Superintelligence Inc., the company co-founded by Ilya Sutskever, says it has "one goal and one product: a safe superintelligence", and plans to advance capabilities "as fast as possible while making sure our safety always remains ahead".
• Microsoft AI proposes "humanist superintelligence", bounded and in service of people. Its CEO Mustafa Suleyman wrote that "we are not building a superintelligence at any cost, with no limits".
• OpenAI wrote in November 2025: "we treat the risks of superintelligent systems as potentially catastrophic" and "no one should deploy superintelligent systems without being able to robustly align and control them". It also raised the question of whether "the whole field should slow development" as systems approach recursive self-improvement.
• The US administration signed a voluntary industry accord on 29 September 2026, according to Nextgov and Al Jazeera, and at the United Nations its science adviser said the US "totally rejects" a global control scheme for superintelligence, according to MeriTalk. See our explainer.

Notice that the building side does not say "no limits". Almost every lab text we read includes a condition: safe, aligned, controllable, bounded. And the prohibition statement is conditional too: "not lifted before" safety and public buy-in. The real disagreement is less "ban or build" than three questions.

• Who decides? Companies and a voluntary accord, national governments, or an international body.
• What proves it is safe? Nobody has an agreed test for superintelligent systems, and the International AI Safety Report notes that evidence on risks is slow to emerge.
• Can one country stop alone? The race argument says no, because a rival would continue. The prohibition argument says that is exactly why an agreement is needed.

• Check which meaning of the word is in playA ban on "superintelligence" and a company's "personal superintelligence" may not describe the same thing. See superintelligence vs AGI.
• Check the status of every billIntroduced is not passed. Most bills in any country never become law.
• Separate present risks from future onesToday's AI has real, present risks that your business can manage now. See our risk guide.
• Follow the money and the mandateBoth campaigners and companies have reasons to raise the stakes. Weigh evidence over volume.