HomeAIAI Policy & SafetyPAC Chair Says AI Is Approaching Government "Like a Tsunami"

PAC Chair Says AI Is Approaching Government “Like a Tsunami”

The chair of the House of Commons Public Accounts Committee (PAC) has warned that artificial intelligence is approaching the UK government “like a tsunami”, and that the civil service may not yet have the skills to cope.

Sir Geoffrey Clifton-Brown makes the warning in his second annual report as PAC chair (HC 629), published on 14 October 2026. The report is his commentary on the committee’s scrutiny of public spending over the past year. It is not law and does not set government policy, but it shows where Parliament’s spending watchdog thinks the biggest risks to taxpayers’ money now sit, and AI is near the top of that list.

“It is approaching government like a tsunami, and I remain unconvinced that the skills, expertise and capability within the civil service that are needed to harness this technological revolution are yet in place,” he writes in the foreword.

Only about 6% of civil servants work in digital and data roles

Later in the report, Clifton-Brown repeats the warning: “AI is approaching government like a tsunami, and I remain sceptical that government is prepared for the enormity of this challenge.”

He says AI offers “significant opportunities” for government, but only if it gets the right specialist skills in place. He adds that it is still unclear how the government’s planned reforms will close “the digital skills gap in practice and by when”. He also warns that repeated reorganisations of Whitehall departments, most recently the creation of the Department for Business, Innovation, Science and Trade, could make things harder.

According to the report, around 6% of civil servants work in digital and data roles, below the levels seen in many high-performing organisations. The government wants to raise that to 10%. Clifton-Brown says this will need a “sustained effort” to recruit, develop and keep specialist staff across the public sector.

Legacy IT is named as a cyber risk for the fourth year running

The report says the government is “showing a worrying lack of urgency in addressing the long-standing issue of legacy IT systems”. It says these old systems create significant cyber security risks and stop departments from using new technology.

The warning builds on the committee’s own earlier work. In its Use of AI in Government report (HC 356), published in March 2025, the PAC said out-of-date legacy technology and poor-quality data were putting AI adoption in the public sector at risk. That report recorded the Department for Science, Innovation and Technology’s estimate that 28% of central government systems met its definition of legacy in 2024, and said 21 of the 72 highest-risk legacy systems prioritised in the 2022–25 digital and data roadmap still lacked remediation funding.

The risk is not abstract. TSN has reported how 32 UK government services were taken offline after the PHP2Shell vulnerability, mostly as a precaution, and how the Legal Aid Agency breach exposed records going back to 1978.

Fraud, benefits and asylum figures in the report

Clifton-Brown also names old IT, along with weak digital skills and leadership and difficulty sharing data, as barriers to tackling fraud. The report covers most of government, starting with defence. Other figures it gives include:

  • Fraud and error: between £55 billion and £81 billion a year.
  • Benefits: the Department for Work and Pensions (DWP) estimates it overpaid £9.9 billion, or 3.2% of benefit spending, through fraud and error in 2025–26. It aims to cut this to 2.8% by 2028–29, which Clifton-Brown says lacks ambition.
  • Asylum: more than 100,000 people claimed asylum in 2025. The Home Office and Ministry of Justice spent around £4.9 billion on asylum in 2024–25.
  • NS&I: the state savings bank’s digital transformation programme, started in 2020, was estimated in December 2024 to cost £3 billion, £1.3 billion more than in 2020. It has been rated “red” since 2022–23.
  • Bank of England: by contrast, the Bank’s new Real-Time Gross Settlement (RTGS) payment system, which settles around £790 billion of sterling payments a day, launched successfully in April 2025. It cost £431 million over nine years. The report holds it up as a model for government.

Why the RTGS contrast matters more than the tsunami line

The “tsunami” phrase will get the headlines, but the more useful point in the report is the comparison between NS&I and the Bank of England. Both are public bodies that set out to replace core systems. One is years late and £1.3 billion over its original estimate; the other replaced a system that moves around £790 billion a day, and had to work first time because the old one was switched off at the same moment. The difference the report draws is in how the risk was managed, not in how new the technology was.

That matters for AI because the government’s AI plans sit on top of exactly the systems the committee keeps flagging. AI tools need clean, shareable data, and much of the state’s data still lives in software that suppliers no longer support. Adding AI to that base without fixing it risks automating the same errors faster, which is the concern behind findings such as the Open Rights Group’s review of Home Office AI asylum summaries.

There is also a fraud angle that is easy to miss. With losses to fraud and error put at £55 billion to £81 billion a year, better data analytics is one of the clearest ways AI could save public money. The report’s complaint is not that government is using AI too much, but that weak skills, poor data sharing and ageing IT are stopping it from using such tools well enough to catch fraud. Seen that way, the warning is as much about missed benefits as about risk.

None of this is binding. In a response to the committee, the government said a Technology Modernisation Action Plan is expected later in 2026. The next tests are whether that plan sets clear funding for the highest-risk legacy systems, and whether the share of digital and data specialists moves measurably towards the 10% target.

Related reading: how UK critical national infrastructure is coping with cyber attacks.

Sources

Related stories

Follow @tsnmediahub on X for breaking AI news as it happens.

Follow on X

Share this story

Latest stories

More in this category

Latest stories

Explore the TSN Atlas: Every topic we cover, mapped. Click a bubble to dive in.