The Government’s use of artificial intelligence (AI) is expanding, but departments must work together to manage risks effectively.
Cabinet Office officials recently told the Treasury Committee that AI tools are used in customer service across many Government departments. However, HMRC confirmed that its helpline advisers do not currently use AI, though it is trialling a chatbot on its website.
HMRC has used machine learning and natural language processing for over a decade in compliance targeting and debt risk prediction. Its most well-known tool, HMRC Connect, was launched in 2010 to tackle tax evasion.
An HMRC insider said all AI initiatives follow an AI assurance, ethics, and risk management framework, which external ethics experts review. HMRC stressed that AI systems impacting taxpayers must be explainable, human-supervised, and compliant with data protection rules.
The professional standards committee, chaired by HMRC’s director general for customer strategy, last met in October 2024. Independent advises have recommended forming an AI steering group to collaborate across departments and avoid a siloed approach.
HMRC’s chatbot recently underwent second-phase testing with 15,000 business users. While initial feedback was positive, concerns remain over accuracy and AI ‘hallucinations’. Users are warned to verify chatbot responses against GOV.uk.
Separately, the Treasury Committee has launched an inquiry into AI use in banking, pensions, and financial services. Submissions are due on 17 March.
Talk to us about your business’s experience with AI.