Taxpayer uses AI to argue against penalty
One taxpayer recently decided to use AI to generate a defence against an HMRC penalty. How did things turn out for her?
Mrs Harber (H) received a penalty from HMRC for failing to notify a capital gains tax liability. She appealed to the First-tier Tribunal on the grounds that she had a “reasonable excuse”; namely that her mental health was poor. What made the appeal interesting was that H used an AI application, similar to ChatGPT, to produce a list of previous cases where an argument of a reasonable excuse due to mental health was accepted.
Upon review of the cases presented by H, the judge found a large number of discrepancies. It turned out that the application had more or less invented the cases, though there were occasional hallmarks of actual decisions present. None of the cases could be located, and the one that had “some” similarity to a real case had the wrong year, which was a case that the taxpayer had lost. The tribunal accepted that H did not know the cases had been created by the application, so there was no question of her trying to mislead deliberately. The case should serve as some relief that the robots aren’t ready to take over just yet!
After all that, the actual decision itself seems relatively boring. H did not convince the tribunal that she had a reasonable excuse for not notifying HMRC of her liability. There would have been nothing to stop her seeking advice from a (human!) tax advisor.
Related Topics
-
HMRC bungles 2026/27 PAYE codes for pensioners
For some pensioners, the 2025/26 winter fuel payment should be collected via their 2026/27 PAYE code. HMRC has started to issue PAYE codes for the new tax year, but the extra charge is missing. What's going on?
-
Are you including too much income in your calculations?
Your business is partly exempt and you claim input tax on your mixed costs and general overheads by using the standard method based on turnover splits. What income should you exclude from the calculations?
-
Electronic VAT return





This website uses both its own and third-party cookies to analyze our services and navigation on our website in order to improve its contents (analytical purposes: measure visits and sources of web traffic). The legal basis is the consent of the user, except in the case of basic cookies, which are essential to navigate this website.