HMRC to use AI from British tech firm to spot fraud and tax return errors

HM Revenue and Customs (HMRC) has agreed a £175 million, 10-year partnership with British AI firm Quantexa to enhance fraud detection and tax return accuracy. The AI technology will analyze HMRC data alongside external sources to uncover hidden fraud networks and unintentional errors, while also supporting customer service teams. Quantexa’s solution will help track misdirected payments and assist in resolving disputes, though automated decisions will require human review. The company emphasizes transparency, stating that AI will ‘support human decision-making’ rather than replace it, with data remaining secure within HMRC’s systems. The deal aligns with the UK government’s push for ‘digital sovereignty,’ reducing reliance on US-based tech providers. It follows concerns over a £330 million NHS AI contract awarded to Palantir, and comes as complaints against HMRC surged to over 93,000 in 2024-25—up from 70,000 in 2020-21—primarily due to slow response times. Quantexa, valued at $2.6 billion, already serves major corporations like HSBC and Vodafone. CEO Vishal Marria assured the BBC that HMRC data will never leave the tax office’s environment, and staff working on the project will operate separately from the rest of the business. The initiative marks another step in the UK’s adoption of AI for public services, following a separate agreement with OpenAI earlier this year.