Department for Work and Pensions (DWP) Algorithm Wrongly Flags 200,000 for Housing Benefit Fraud

The Department for Work and Pensions (DWP) deployed an automated algorithm to identify potentially fraudulent housing benefit claims by analyzing claimants' personal characteristics including age, gender, number of children, and tenancy agreement type. The system does not use artificial intelligence or machine learning but was designed to flag high-risk claims for manual review by council staff. Over a three-year period from 2020-2023, the algorithm flagged claims as potentially fraudulent, but subsequent investigations revealed that only 34-37% of flagged cases actually involved fraud or error, compared to the 64% fraud rate identified during the pilot phase. This meant that approximately 200,000 people were wrongly investigated for housing benefit fraud. The system cost £4.4 million in investigation expenses that did not result in savings. Despite the poor accuracy, the DWP reported that every pound spent on case reviews returned £2.71 in savings according to 2021/22 figures. The data was obtained by civil liberties group Big Brother Watch through Freedom of Information requests. The DWP has expanded its use of AI systems across other benefit programs including universal credit despite warnings about algorithmic bias against vulnerable claimants.