The Chancellor's pre-Budget AI package offers NHS chest X-ray software as its evidence that AI already delivers better public services. A five-hospital randomised trial found no significant difference in how fast lung cancer was diagnosed.
Estimated reading time: 7 minutes
.jpg){kind=link}
In short. HM Treasury's 24 October 2026 announcement, previewing AI measures ahead of next week's Budget, says "AI is already helping to deliver better public services across the country," offering NHS chest X-ray software as its example: "over 210,000 chest X-rays each month... an 11.5% increase in suspected lung cancer cases identified within 24 hours... around 400 more patients receive faster clarity each month, while reducing the average wait for follow-up CT scans from 6.3 to 5.6 days" [1]. None of those figures carries a citation. The largest randomised trial of exactly this technology, LungIMPACT, ran across five NHS hospitals, covered 93,326 chest X-rays and 558 lung cancer patients, and found "no significant reduction in the time from X-ray to cancer diagnosis": a median of 44 days with AI prioritisation against 46 days without, a difference its own researchers call not significant [2]. The same package sets aside 10 million pounds a year "to lead the UK's work in understanding how AI will reshape jobs," two days after Barclays told investors it was cutting 5,000 jobs, about 5% of its workforce, citing "technology change efficiency driven by AI" among its reasons [3]. Three documented mistakes: four stars.
Chancellor of the Exchequer John Healey is due to put AI into his Budget next Wednesday. HM Treasury's preview, published two days ago under the headline "Chancellor unveils Britain's ambitious plan for AI," lists a package of measures worth more than 600 million pounds: a new AI Biosecurity Lab, a bigger AI Security Institute, a National Cyber Shield, a Sovereign AI Fund top-up, chip design training, and more [1]. techUK's chief executive calls it "a strong £600m+ package" [1], and the sums in the press release add up to roughly £621 million, so that figure checks out. The document also makes a case for urgency: AI is not a future prospect, it argues, but something "already helping to deliver better public services across the country" right now. The two pieces of evidence it chooses to make that case, and the one commitment it makes to the people most worried about AI, do not hold up as well as the headline figure does.
The claims, tested
| The claim | What we found | Verdict |
|---|---|---|
| "In the NHS it is already being used to help clinicians analyse over 210,000 chest X-rays each month, around a third of all chest X-rays in the NHS. This has led to an 11.5% increase in suspected lung cancer cases identified within 24 hours, meaning around 400 more patients receive faster clarity each month, while reducing the average wait for follow-up CT scans from 6.3 to 5.6 days" [1], offered as the lead example of AI "already helping to deliver better public services" | LungIMPACT, the largest randomised trial of AI chest X-ray prioritisation, ran across five NHS trusts, 93,326 X-rays and 558 cancer diagnoses. AI cut the median time for a radiologist to report an X-ray from 47 hours to 34 hours, but "this improvement did not cascade into faster progress through the rest of the diagnostic pathway": median time from X-ray to cancer diagnosis was 44 days with AI against 46 days without, "a two-day difference that was deemed not significant" [2] | The press release's own figures carry no source and describe flagging speed and scan waits, not the outcome that matters to a patient. The one study that measured that outcome directly, at comparable NHS scale, found AI prioritisation made no significant difference to it |
| "£10 million per year to lead the UK's work in understanding how AI will reshape jobs and industry, and how Britain needs to prepare for it, addressing concerns about AI's replacement of jobs from workers across the country" [1] | Two days before this press release, Barclays' third-quarter results said the bank would cut 5,000 jobs, about 5% of its roughly 93,000-strong global workforce, with 60% of the cuts in support functions including developer roles, and named "organisational simplification and technology change efficiency driven by AI" as a cause [3] | The document treats AI's effect on jobs as a question to be studied over time. The public record, two days old when this was published, shows a single UK bank already attributing thousands of live job cuts to AI efficiency |
| The same press release names two separate funding lines for studying AI's effect on employment: the £10 million a year "to lead the UK's work in understanding how AI will reshape jobs and industry" [1], and "an additional £23 million over the next three years for the AI Economics Institute... the Institute's comprehensive research agenda on how AI is affecting productivity, growth, jobs and living standards" [1] | The document does not say how the two relate, which one leads, or why a "world's first" institution already funded to research AI's jobs impact needed a second, separately-funded work-stream announced in the same breath for what reads as the same question | Two commitments to answer the same question, in the same document, with no stated division of labour between them |
The mistakes, counted
The document's headline evidence for AI already working is contradicted by the best available trial of the same technology (1). The 24 October press release opens its case for urgency with a single concrete example: NHS chest X-ray software [1]. LungIMPACT, published in Nature Medicine on 24 March 2026 and run by UCLH with UCL and the University of Nottingham, is described by its own authors as "the largest randomised trial of its kind" [2]. Its principal investigator, UCLH consultant radiographer Nick Woznitza, put the finding plainly: "The hypothesis was that getting the X-ray reported faster would cascade into a faster diagnosis, but the NHS system simply could not respond at the same pace. The bottleneck isn't the reporting; it's everything that happens next: telling the patient, the CT appointment, the clinic slot, the multidisciplinary meeting" [2]. The press release's own statistics, uncited, describe exactly the stage the trial found does not matter, flagging speed and scan waits, not whether a patient is diagnosed any sooner.
The jobs-impact fund is framed as addressing a future concern that is already a current fact (2). Barclays' third-quarter results, published two days before this announcement, attributed part of 5,000 planned job cuts directly to AI-driven efficiency [3]. The press release's own words describe "concerns about AI's replacement of jobs from workers across the country" as something a new 10-million-pound-a-year work-stream will help government "lead the UK's work in understanding" [1]. Understanding is not nothing, but the document gives no date by which that understanding will exist, and no sign that the week's own financial news informed the figure.
Two funding lines for the same question, unreconciled (3). The AI Economics Institute, already running since June 2026 under Nobel laureate Simon Johnson, exists specifically to research "how AI is affecting productivity, growth, jobs and living standards," now with an extra £23 million over three years [1]. The same document, in the same breath, creates a separate £10-million-a-year commitment "to lead the UK's work" on how AI reshapes jobs [1]. A reader is given no way to tell whether these are one effort described twice, two efforts that overlap, or two efforts with a clear division nobody wrote down.
Credit where due
The headline total holds up: techUK's "£600m+" framing of the package is accurate arithmetic, not spin, against the roughly £621 million the press release itself lists [1]. The document is also not pure cheerleading throughout: CEPI's deputy chief executive, quoted in the same release on the AI Biosecurity Lab, is candid that "rapid advances in AI are creating new risks related to the design and modification of pathogens, increasing the potential for both accidental and deliberate harm," before noting the same technology's upside, a more balanced framing than the ministerial quotes around it [1]. And the AI Economics Institute itself, whatever its overlap with the new jobs fund, is a real and independently reported piece of infrastructure, led by a credible outside economist rather than run entirely in-house [4].
Verdict
Four stars, from three documented mistakes. The arithmetic behind the headline figure is sound, and some of the package, the Economics Institute chief among it, is real capacity rather than announcement for its own sake. But the one piece of evidence the document offers that AI is already working in public services is contradicted by the best trial anyone has run on it, and the one commitment aimed at workers worried about their jobs reads as a promise to start understanding a problem the week's own financial news shows is already under way.
The star score counts three documented mistakes. First, the press release's uncited NHS chest X-ray statistics (an 11.5% increase in suspected lung cancer cases flagged within 24 hours, a CT wait cut from 6.3 to 5.6 days) are offered as evidence that AI is "already helping to deliver better public services", while the largest randomised trial of the same technology, LungIMPACT (93,326 X-rays, five NHS hospitals, published 24 March 2026), found no significant difference in time to actual cancer diagnosis. Second, the £10 million a year earmarked "to lead the UK's work in understanding how AI will reshape jobs" is framed against "concerns about AI's replacement of jobs", two days after Barclays attributed part of 5,000 planned job cuts to AI-driven efficiency, with no date set for when that understanding will arrive. Third, that same £10 million a year is not reconciled against the AI Economics Institute's own, separately funded remit to research "how AI is affecting... jobs", with no explanation of how the two relate. Three falls in the 1 to 3 band: four stars; the bands are on the ratings page. This piece makes no finding about the conduct of any named minister, official or company executive; the press release's chosen evidence and funding structure are institutional facts, not any one person's decision. Checked directly against HM Treasury's press release, the LungIMPACT trial's published findings and Barclays' third-quarter results, all fetched 25 October 2026.
Sources
- HM Treasury, "Chancellor unveils Britain's ambitious plan for AI" (24 October 2026). https://www.gov.uk/government/news/chancellor-unveils-britains-ambitious-plan-for-ai
- UCLH NIHR Biomedical Research Centre, "AI-driven prioritisation of abnormal chest X-rays does not speed up lung cancer diagnosis", reporting the LungIMPACT trial published in Nature Medicine, 24 March 2026. https://www.uclhospitals.brc.nihr.ac.uk/news/ai-driven-prioritisation-abnormal-chest-x-rays-does-not-speed-lung-cancer-diagnosis
- CNBC, "Barclays to slash 5,000 jobs in AI efficiency push" (22 October 2026). https://www.cnbc.com/2026/10/22/barclays-ai-jobs-cuts.html
- Inomics, "UK Government Launches AI Economics Institute" (June 2026). https://inomics.com/news/uk-government-launches-ai-economics-institute-1554110