HMRC could force UK taxpayers to hand over their artificial intelligence (AI) chatbot histories under new rules designed to catch tax dodgers. The tax authority's powers to obtain computer records to ensure Britons have paid enough tax could be expanded under the Budget Bill, which will become law at the end of May.
Proposals in the bill could see HMRC demand taxpayer information from third parties including AI platforms, as well as requiring taxpayers themselves to supply personal data. Those who do not comply could face fines of £300, with taxpayers who fail to report inaccuracies potentially liable for up to £3,000, The Telegraph reports.
Privacy Concerns Raised
Anne Strickland, of the Taxpayers' Alliance campaign group, warned that demanding private chatbot data "would be a deeply intrusive overreach by the taxman". She added: "HMRC already has sweeping powers, and expanding them further risks trampling over the privacy of ordinary Britons."
Seb Maley, chief executive of tax insurance provider Qdos, said HMRC was making sure it "isn't left behind", as a growing number of people and businesses use AI to manage their tax affairs. The tax authority can only issue "Schedule 36" notices, demanding information from companies or individuals, if they can show a "reasonable" need to do so.
New Debt Recovery Powers
The move comes after Chancellor John Healey handed the tax authority new powers allowing it to recover debts directly from bank accounts from July 2029. The crackdown will target Brits who refuse to engage with the taxman despite possessing sufficient funds.
Under the new rules, HMRC will be able to automatically deduct debts of up to £5,000 for current and savings accounts and up to £10,000 for companies. The renewed efforts to recover unpaid tax are part of Labour's ongoing move to claim an estimated £59.2billion owed in the 2024/2025 tax year.
HMRC Response
A spokesperson for HMRC said: "This is about updating our existing information powers to reflect how tax information is now stored and processed digitally. It is not about going through everyone's use of large language models. We can only request information where we already have a lawful basis to do so."
