DWP deploys machine learning model for UC advances

The Department for Work and Pensions (DWP) has used a machine learning model to flag potentially fraudulent claims for Universal Credit (UC) advances since May 2022, the National Audit Office (NAO) has reported¹. The model is one of a set of counter-fraud tools described in the NAO's October 2025 report on benefit overpayments due to fraud and error¹.

The NAO said DWP has deployed one machine learning model, with four others in development and testing¹. The report also sets out the wider counter-fraud programme: it began in February 2022 with seven agents and by April 2024 involved 3,100 DWP staff¹. DWP has been awarded £6.7 billion of earmarked funding for fraud and error activity covering 2020-21 to 2028-29, and the funding is backloaded, with 52% of the total (£3.5 billion) due in the three years from 2026-27¹.

The NAO reported that DWP saved an estimated £4.5 billion in Annually Managed Expenditure through counter-fraud activities from April 2022 to March 2025¹. It achieved savings of £1.35 billion in 2023-24 against a target of £1.3 billion, and £2.0 billion in 2024-25 against a target of £1.7 billion¹. In July 2025, DWP published detailed information on its fairness analysis for the first time¹.

On overpayments, the NAO estimated that 3.3% of benefit expenditure was overpaid in 2024-25, down from 3.6% in 2023-24, or £9.5 billion against £9.7 billion¹. The UC overpayment rate fell by 2.7 percentage points, from 12.4% in 2023-24 to 9.7% in 2024-25¹. UC accounted for 67% of overpayments by value in 2024-25, and for the first time since it was rolled out nationally in 2018 it did not have the highest overpayment rate across all benefit lines; Pension Credit had the highest estimated rate at 10.3%¹.

Why it matters for households

UC advances are payments made against a future UC award, and the guide to advances and budgeting loans explains how they work. The NAO's report concerns claims flagged as potentially fraudulent, not confirmed fraud, and it does not set out how many claims were flagged by the model or what happened to them; those figures have not been reported¹.

The overpayment figures describe money paid that claimants were not entitled to, and DWP can recover overpayments. The NAO reported that the main causes of UC overpayments in 2024-25 were claimants, mainly self-employed claimants, not declaring income from work in full, claimants failing to declare that they lived with a partner, and claimants not declaring all their financial assets¹. The report covers overpayments only, not underpayments¹.

DWP makes welfare payments to more than 23 million people across Great Britain, and paid £290.8 billion in benefits including State Pension in 2024-25, with £7.3 billion spent on running costs¹. The NAO said DWP rated as "red" the risk that its plans to reduce fraud and error are not successfully executed or cannot mitigate the increased propensity for fraud in society¹. Successive Comptroller and Auditor Generals have qualified their audit opinions on the regularity of DWP's accounts, excluding State Pension, for 37 years because of material fraud and error¹.

What happens next

The government introduced the Public Authorities (Fraud, Error and Recovery) Bill to Parliament in January 2025¹. DWP's refreshed fraud and error strategy was approved in November 2024, and at the time of the NAO's work DWP had started to develop implementation and evaluation plans¹. The Spring Statement 2025 forecast that overpayments would fall to the pre-pandemic level of 3.1% by 2028-29¹.