The British government’s effort to expand its use of artificial intelligence is colliding with a more basic technology problem: too many aging systems and too few people equipped to modernize them. The Register reported on October 14 that Sir Geoffrey Clifton-Brown, chair of the House of Commons Public Accounts Committee, used his second annual report to warn that Whitehall is moving too slowly on longstanding digital weaknesses even as AI creates new opportunities for criminals and hostile states.
The warning matters because it reframes government AI adoption as an infrastructure and staffing challenge, not simply a contest to acquire new models. According to The Register’s account, Clifton-Brown said ministers have announced reforms intended to strengthen digital capability but have not explained how, or when, they will close the civil service skills gap. He also expressed concern that another departmental reorganization could make the problem harder to manage.
The staffing figures cited in the report show the size of that gap. Only 6 percent of civil servants work in digital and data roles, compared with the government’s own target of 10 percent. The committee said senior specialists are particularly scarce and that hitting the target will require the government to recruit and retain substantially more people with the necessary expertise.
Legacy technology is the other half of the problem. The committee’s report said the government is showing a lack of urgency in dealing with old IT systems that create cybersecurity risks and make it harder to benefit from newer technology. That creates an awkward sequencing problem: adding AI capabilities may increase complexity while the underlying systems remain difficult to secure, maintain or connect.
The watchdog linked that weakness to a wider fraud problem. The Register reported that fraud and error cost British taxpayers an estimated £55 billion to £81 billion each year, based on figures cited by the committee. Generative AI could give fraudsters more sophisticated tools, the report warned, while departments that lack modern systems and specialist staff may struggle to keep pace.
One example came from the Department for Work and Pensions, which estimated that it overpaid £9.9 billion in benefits through fraud and error during 2025-26, equal to 3.2 percent of its spending. The department aims to reduce that rate to 2.8 percent by 2028-29. Clifton-Brown argued that even if the target is met, the remaining losses would still amount to billions of pounds.
The report also pointed to troubled modernization work as evidence that large government technology programs can become expensive without producing predictable results. National Savings & Investments began overhauling its IT in 2020, but by December 2024 the projected cost had reached £3 billion, or £1.3 billion more than originally expected. The project had been rated red since 2022-23, and NS&I could not provide lawmakers with a current spending figure or a realistic completion date.
A contrasting case suggests that failure is not inevitable. The Bank of England completed an overhaul of its Real-Time Gross Settlement system in April 2025 for £431 million. The committee directed its recommendations toward Whitehall rather than the Bank, signaling that the central bank’s approach may offer practical lessons for departments still wrestling with older systems.
The central message is that AI policy cannot be separated from the condition of the machinery beneath it. The committee’s assessment does not say the UK should abandon AI. It says, in effect, that ambitious deployments will remain exposed unless government first strengthens its digital workforce, replaces risky legacy systems and proves it can deliver modernization projects with credible budgets and timelines.
Comments
Loading comments…