Seventy Signatures Against the Singularity: UK Lawmakers Demand the World’s First Superintelligence Ban

On September 11, 2026, a letter landed on the desk of Prime Minister Andy Burnham carrying an unusually blunt demand for Westminster: ban the creation of artificial superintelligence, and lead an international coalition to do the same. The signatories — more than 70 MPs and members of the House of Lords, drawn from across party lines — were not fringe backbenchers. Their number included names attached to nearly every serious AI security intervention the UK parliament has debated this year, and their intervention followed a Labour backbencher’s ten-minute rule bill that had already made legislative history.

Five days earlier, on September 8, Alex Sobel MP stood up in the House of Commons and introduced the Artificial Superintelligence Security Bill — the first piece of legislation anywhere in the world whose explicit purpose is to prohibit the development, deployment and operation of artificial superintelligence (ASI) systems. Within seventy-two hours, the campaign group ControlAI had helped convert that lonely ten-minute-rule motion into a cross-party letter with seventy-plus parliamentary signatures. It is the fastest an AI-safety proposal has ever moved from a single MP’s presentation bill to a formal appeal to a sitting prime minister.

What the bill actually does

The text, now visible on the parliamentary bills website, does two things that no legislature has previously attempted in combination.

First, it prohibits. The bill would make it unlawful to develop, deploy or operate systems whose capabilities exceed human performance across effectively all cognitive domains — the working definition of superintelligence used by safety researchers. Second, it reaches backwards down the capability ladder: the bill would empower the government to monitor and, where necessary, restrict “precursors” to superintelligent AI. During his presentation in the Commons chamber, Sobel was explicit about why that second power matters. “No company, government or individual knows how to keep superintelligent AI under human control,” he told the chamber, arguing that waiting until a system displays dangerous capabilities before regulating it is a category error — by then, the moment for effective control has already passed.

The enforcement architecture remains deliberately vague in the bill’s current form, which is standard for a ten-minute rule presentation bill. What matters constitutionally is the principle it establishes: that a line exists, that systems above it are unlawful, and that a democratic legislature — not a company’s safety team — decides where the line sits.

The letter: what seventy signatures are asking for

The Burnham letter, first reported by The Guardian, goes further than the bill itself in one respect: it asks the prime minister to treat superintelligence as a global coordination problem, not merely a domestic regulatory one. The signatories urge Burnham to back a ban on ASI creation and to lead an international effort to establish the same prohibition among allies — an implicit invitation to revive the UK’s AI Safety Summit diplomacy, but pointed at a far more categorical target than anything discussed at Bletchley Park in 2023.

The Express followed up with a breakdown of the signatory list, reporting cross-party names and noting that the letter explicitly frames the request as a matter of preventing AI-caused human extinction. The letter’s most quoted line, according to multiple reports, leans on the numbers coming out of frontier labs themselves: warnings that senior figures inside AI companies privately assign a greater-than-ten-percent probability to AI killing everyone.

Why now: the Coxon signal

The proximate trigger is a name that has dominated UK AI coverage for a week: Jacob Coxon. The former Anthropic employee — now the most prominent insider to break ranks — has spent the days around the letter’s delivery repeating a single claim in broadcast interviews: that people inside frontier labs believe the technology they are building could kill us all by the end of the decade, and that this belief is more widespread among practitioners than among the executives who set deployment schedules.

Coxon’s intervention mattered because of who he is, not what he said — similar warnings have come from Yoshua Bengio and a long line of safety researchers. But an insider accusation carries a different evidentiary weight in a parliamentary office. Several outlets covering the letter noted that the Anthropic-employee claim was the hook on which the seventy signatures were hung: a former employee of a leading lab asserting that insiders themselves believe the technology could wipe out humans by decade’s end.

The timing also coincided with Burnham’s own public positioning. Days before the letter, the prime minister acknowledged in an interview that AI poses risks to national security while insisting it also offers solutions for keeping the country safer — the classic dual-use framing that AI-skeptical MPs read as hedging. The letter’s signatories are, in effect, calling that hedge: if the PM accepts the risk, they argue, the only proportionate response to extinction-level capability is prohibition, not monitoring.

A transatlantic echo

Westminster is not moving alone. Three days before Sobel’s bill was tabled, Senator Bernie Sanders and Representative Greg Casar announced the Ban Artificial Superintelligence Act in the United States Congress — legislation that would permanently prohibit ASI development and temporarily pause advanced AI work below that threshold. The near-simultaneous appearance of ASI-ban bills in both legislatures, promoted by overlapping campaign networks, looks less like coincidence and more like a coordinated test of whether categorical prohibition is now politically speakable in major democracies.

Neither bill is likely to become law in its current form. Both are presentation bills and private members’ vehicles — instruments for agenda-setting, not statute-writing. But the trajectory matters. A year ago, “ban superintelligence” was a position confined to campaigners and a handful of researchers. Today it has seventy parliamentary signatures in the UK, a Senate sponsor in the US, and a prime minister publicly acknowledging the underlying risk assessment.

The obvious objection — and the campaign’s answer

Critics of the bill fall into two camps. The first argues the definitional problem is fatal: “superintelligence” is not a measurable threshold, and any statute keyed to it will either capture nothing or everything. The second argues prohibition is strategically naive — that if the UK refuses to build, less scrupulous actors will not, and Britain will have traded influence over the technology’s development for the moral satisfaction of having renounced it.

ControlAI’s answer to the first objection is the precursor-monitoring power, which shifts the regulatory trigger from an unmeasurable endpoint to observable capability milestones. Its answer to the second is historical analogy: the same coordination argument was made against nuclear non-proliferation treaties, bans on human germline editing, and chemical weapons conventions — regimes that are imperfect but demonstrably shaped state behaviour compared with an unregulated counterfactual.

The bill’s second reading has not yet been scheduled. Under the standard arithmetic of private members’ legislation, the odds of it advancing without government support are close to zero. But the seventy-signature letter changes what the government must visibly refuse. For the first time, a British prime minister is being asked, on the record, to say whether his country should be permitted to build a machine smarter than every person in it — and a growing cross-party bloc has committed to remembering the answer.