The Cabinet Office has revealed that government anti-fraud teams recovered a record £480m in the 12 months from April 2024 — the largest sum ever clawed back in a single year.
The savings were achieved through a combination of cross-checking data across departments and deploying a newly developed AI tool. Ministers say the money will now help fund frontline services, including the recruitment of nurses, teachers and police officers.
Of the total recovered, £186m was linked to Covid-related fraud. While ministers have repeatedly pledged to recover losses from the pandemic, the sum remains a small portion of the more than £7bn in missing funds Labour claimed before last year’s general election.
The crackdown has also prevented hundreds of thousands of businesses with suspicious Bounce Back Loans from dissolving before repaying their debts. The loan scheme, introduced during the pandemic to support small businesses with up to £50,000, has faced heavy criticism for weak safeguards that left it open to widespread abuse. In one case uncovered, a woman fabricated a company and funnelled the loan money overseas to Poland.
Cabinet Office minister Josh Simons will outline the savings at an international anti-fraud summit on Wednesday, held jointly with the US, Canada and Australia. He said the government is now using “cutting-edge AI and data tools” to protect taxpayers’ money rather than “line the pockets of scammers and swindlers.”
At the heart of the initiative is the Fraud Risk Assessment Accelerator, an AI system designed within the Cabinet Office to detect vulnerabilities in new policies before they can be exploited. Officials believe the tool could make future policies “fraud-proof” and confirmed it will now be rolled out across other departments.
The UK government also plans to license the system internationally, with the US, Canada, Australia and New Zealand expected to adopt it. However, the move is likely to spark fresh debate among campaign groups critical of government reliance on AI.
Concerns about algorithmic bias persist after a separate welfare fraud detection system was found last year to discriminate based on age, disability, marital status and nationality.
