For the past two years, whenever the British government wanted to talk about artificial intelligence, it reached for one number: £400 billion. That was the promised economic boost AI would deliver to the UK, a figure rolled out again and again to reassure a nervous public. Every time fears surfaced about mass unemployment, about algorithms replacing skilled workers, about entire industries being disrupted overnight, ministers would point to that gleaming total. It sounded like a national windfall, a prize for embracing the future. The message was simple: don’t worry about the risks of AI, because the rewards are enormous. But last week, that carefully constructed confidence cracked. The UK’s AI minister, Kanishka Narayan, admitted in response to a parliamentary question that the government’s own estimate of AI’s economic benefit is far smaller than the £400 billion figure it has been quoting. In fact, the government’s central estimate is now just £120 billion, and the full range is between £55 billion and £140 billion by 2030. That means the headline number, the one used in official speeches, government documents, and policy debates, was more than three times higher than the government’s own best guess. The admission didn’t come with a dramatic announcement or a press conference. It slipped out through a written parliamentary answer, the kind of routine bureaucratic disclosure that rarely makes news. But it matters, because it reveals how a single commissioned report, paid for by one of the world’s most powerful technology companies, became enshrined as official government truth.
The source of that £400 billion figure is now clear, and it’s not a comfortable story for those who defended it. According to Narayan, the number was drawn from a 2023 report commissioned by Google and produced by Public First, a British consultancy now known as PF Global. That report was publicly available, but it was never an independent assessment. It was corporate research, paid for by a company with an enormous stake in the expansion of AI. Yet government departments and agencies treated it as a reliable projection for nearly two years, under two different governments. What makes this particularly troubling is that the government had its own lower estimate long before this week’s admission. Ministers have apparently been aware of the more modest figures since at least November 2025, almost a year before Narayan’s recent statement. Even so, government bodies kept repeating the old £400 billion claim. Skills England, a government agency responsible for apprenticeships and skills policy, was still publishing the inflated figure four months after the lower estimate was supposedly in circulation. This is not just a small bureaucratic slip. It suggests that the £400 billion number was not carefully checked, because it was convenient. It served a purpose. It helped sell a story about AI as a national opportunity rather than a threat. And in doing so, it made honest public debate about the real trade-offs of AI much harder.
Narayan’s parliamentary answer is notable for another reason: it is the first recorded evidence, according to WIRED, that the government has set the two numbers side by side. It is also the first time it explicitly acknowledged that the £400 billion figure came from a report commissioned by Google. Yet the answer did not withdraw the £400 billion figure, nor did it say that figure had been wrong. It simply presented the new, lower estimate as a “more recent assessment” while adding that even this was “hugely uncertain.” That phrase matters. The government is not merely correcting an outdated number; it is admitting that its ability to predict the economic impact of AI is shaky at best. That is a far more honest position than repeating a flashy headline, but it also creates a problem. If the government cannot say with confidence how much AI will help the economy, then it cannot easily justify policy choices based on that expected windfall. It also cannot fully answer the public’s deeper questions. Will AI really create more jobs than it destroys? Will the gains be shared widely or captured by a few large companies? Will ordinary workers see better wages or more precarious employment? The £400 billion figure, whatever its flaws, offered a comforting answer to all of those questions. Its collapse leaves those anxieties unresolved.
PF Global, the consultancy behind the original Google-commissioned report, has offered its own explanation for the gap between the two numbers. Jonathan Dupont, the firm’s head of economics, said the difference was partly because the government has been slower to adopt AI than the model assumed. He pointed out that AI is moving very quickly and that, as more information emerges, models must be updated. In his words, compared to the older model, the numbers the government uses assume a much slower rate of adoption than looked possible back in 2023. That sounds reasonable, but it also reveals something important about how these projections work. The original £400 billion figure was never a precise forecast. It was an estimate built on assumptions, and those assumptions were optimistic. They assumed rapid adoption, smooth implementation, and significant productivity gains. When those assumptions are adjusted, the number falls by more than half. That should have been acknowledged much earlier. Instead, the original figure was allowed to become a kind of official slogan. It was repeated so often that it took on a life of its own, untethered from the evidence behind it. This is a common pattern in technology policy, where compelling narratives often matter more than careful analysis. But when the narrative collapses, trust in the underlying policy suffers too.
The political reaction to this revelation has been pointed. Dame Chi Onwurah, the Labour MP who chairs the Commons Science, Innovation and Technology Committee, told WIRED that the government is right to promote the potential of technology like AI to improve public services. But she went on to identify a central problem that her committee has repeatedly found: hype. That single word captures a lot. Hype is not the same as deception. It is the tendency to exaggerate benefits, downplay risks, and treat uncertain projections as established facts. Hype can be generated by companies that profit from public enthusiasm, by politicians who want to sound forward-looking, and by officials who want to avoid difficult questions. The result is a policy environment where expectations are set impossibly high, and where the real, incremental, messy work of digital transformation gets ignored. AI could indeed improve public services. It could help doctors diagnose diseases more quickly, help teachers personalise lessons, help civil servants process paperwork faster. But none of that happens automatically. It requires investment in infrastructure, training, data standards, and safeguards. It requires honest assessment of what works and what doesn’t. And it requires a government willing to tell the public that the benefits will take time and may be modest at first. That kind of honesty is much harder than repeating a big number. It also earns more trust.
In the end, this story is about more than one statistic. It is about how governments make decisions in an age of rapid technological change, and about who gets to shape the stories we tell about the future. The £400 billion figure was not simply a forecasting error. It was a product of a system that rewards optimism, that listens closely to powerful tech companies, and that often prefers a good story to a complicated truth. Google, of course, is not an objective observer. It is a company that stands to benefit enormously from increased AI adoption, and commissioning a flattering report is a familiar part of corporate influence. Public First, now PF Global, was paid for that work, and its head of economics is now explaining away the gap between his firm’s old report and the government’s new numbers. That is the way influence works in modern politics, not through scandal or bribery, but through the careful production of convenient facts. The lesson of this episode is not that AI is worthless, or that it cannot bring real economic benefits. It is that the public deserves better than inflated promises. If the government genuinely wants to build a productive, humane AI economy, it must start with honesty. It must admit uncertainty, explain trade-offs, and resist the temptation to use technology companies as its source of truth. The £400 billion figure may be gone, but the questions behind it remain. And they deserve answers that are as grounded in reality as possible.
