On 22 October 2025, the Future of Life Institute released a public letter titled the Statement on Superintelligence. Unlike the sprawling manifestos that usually accompany moments like this, it is a single sentence:
"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."
Statement on Superintelligence, Future of Life Institute · October 2025
That sentence is doing a great deal of careful work, and it is worth reading twice. It does not say superintelligence should never be built. It does not condemn AI as such. It sets a condition — two conditions, in fact — and asks that development stop until they are met. To understand why this modest-sounding request represents a genuine shift in the debate, you have to understand both what it demands and who was willing to put their name to it.
What the statement actually asks for
The phrasing is precise. Three features distinguish it from earlier open letters, including the well-known 2023 letter calling for a six-month pause on the largest training runs.
- It targets superintelligence specifically. Not narrow AI, not chatbots, not the AI tools already woven into daily life. The prohibition applies to systems that would exceed human capability across essentially all cognitive domains — the category we describe in our explainer on artificial superintelligence. This precision blunts the usual objection that safety advocates want to halt beneficial technology. They do not. They want to halt one specific, and specifically dangerous, frontier.
- It is a conditional prohibition, not a permanent ban. The pause lifts the moment two thresholds are crossed. This reframes the entire question. The default today is: build superintelligence, and establish safety afterward if we can. The statement inverts it: establish safety first, then build. The burden of proof moves onto the developers, where — for a technology of this consequence — it arguably belongs.
- It requires public buy-in, not just expert sign-off. The second condition is democratic, not technical. Even if scientists someday agree superintelligence can be controlled, the statement insists the public must consent to building it. That is a direct challenge to the current arrangement, in which a handful of private labs make a civilisation-altering decision on everyone's behalf — the concern we examine in who controls superintelligence.
Why the signatories matter more than the sentence
Open letters are cheap. What made this one land was the coalition. More than 850 people signed, and the list is remarkable not for its length but for its range.
At the top were the scientists who built the field. Yoshua Bengio and Geoffrey Hinton — both Turing Award laureates, both routinely called godfathers of modern AI — signed, as did Berkeley's Stuart Russell, author of the standard AI textbook. When the people most responsible for a technology's existence ask the world to stop before building its most powerful form, that is not activism. That is testimony.
Then came the builders and the public figures. Apple co-founder Steve Wozniak and Virgin founder Richard Branson. Historian Yuval Noah Harari, actor Stephen Fry, and musician will.i.am. Retired national-security leaders including former Joint Chiefs chairman Admiral Mike Mullen and former National Security Advisor Susan Rice. And — most tellingly — figures who agree on almost nothing else: Prince Harry and Meghan alongside Steve Bannon and Glenn Beck.
That last detail is the whole argument in miniature. When a British royal couple and a MAGA media strategist sign the same sentence, the issue has escaped the usual ideological sorting. Concern about superintelligence is not left or right, not establishment or populist. It is one of the vanishingly few positions in 2025–2026 with genuine cross-partisan reach — which is exactly the kind of coalition that makes binding policy possible.
What the statement is — and what it is not
It is important to be clear-eyed about the limits. A public statement prohibits nothing. It commands no regulator, binds no company, and constrains no government. Anyone reading it as a law would be mistaken.
What it does is move the Overton window — the range of positions considered mainstream and sayable. Two years ago, calling for a prohibition on a class of AI would have been dismissed as fringe. After October 2025, it is a position held publicly by Nobel-tier scientists, iconic entrepreneurs, and political figures across the spectrum. That shift in what is respectable to say is the precondition for everything that follows. Politicians take positions that are already popular; they rarely lead. The statement's function is to make prohibition popular enough that leaders can act on it.
This is precisely how earlier technology-governance movements began. The Pugwash conferences spent two decades building the intellectual and moral consensus that eventually underwrote nuclear arms control. Statements and declarations are not the finish line. They are how the runway gets built.
From a sentence to a treaty
The gap between the statement and reality is the gap the Nakada Foundation exists to close. A declaration establishes that prohibition is desirable and legitimate. It does not establish the machinery to enforce it. That requires three things the statement deliberately leaves unspecified:
- A definition with teeth. "Superintelligence" must be translated into criteria a treaty can act on — compute thresholds, capability evaluations, or both. We examine this in the diplomatic challenge of defining dangerous AI.
- Verification. A prohibition no one can check is the Biological Weapons Convention — a noble ban violated for decades without detection. Credible governance needs monitoring, modelled on the nuclear verification regime.
- Binding commitment. Voluntary pledges have a track record, and it is not encouraging — the subject of our piece on why voluntary commitments fall short. The statement's principle has to become enforceable law.
The Statement on Superintelligence proved something valuable: the will exists, and it crosses every line that usually divides people. Our plan is about converting that will into the compute controls, monitoring agency, and international treaty that would give the statement's single sentence the force it needs. If you agree with that sentence, the most useful thing you can do next is help build the coalition that turns it into policy.
Common questions.
It is a one-sentence public letter released by the Future of Life Institute in October 2025, calling for a prohibition on developing superintelligence until there is broad scientific consensus that it can be done safely and controllably, plus strong public buy-in. More than 850 people signed it, including leading AI scientists, technology founders, and public figures across the political spectrum.
No. It targets superintelligence specifically — systems that would exceed human ability across nearly all cognitive domains — not narrow AI or the everyday AI tools already in use. And it is a conditional prohibition, not a permanent one: development would resume once safety and public consent are established.
The 2023 letter asked for a temporary six-month pause on training the largest models. The 2025 Statement on Superintelligence goes further and narrower: it asks for an open-ended prohibition on superintelligence specifically, lifted only when strict safety and democratic conditions are met. It also attracted a notably broader, more cross-partisan set of signatories.
Not by itself — it is a declaration of principle, not a law or treaty, so it has no enforcement power. Its value is in shifting what is politically mainstream and building the consensus needed for binding rules. Turning the principle into an enforceable reality requires domestic legislation and an international treaty with real verification.