ARTICLE
26 September 2025

The Age Of AI Advisers? – The Pitfalls Of Relying On AI For Legal Advice

AG
Addleshaw Goddard

Contributor

Addleshaw Goddard is an international law firm, almost 250 years in the making. We're trusted by over 5000 organisations, including 50 FTSE 100 companies, to solve problems, deliver deals, defend rights, comply with regulations and mitigate risk. Our work spans more than 50 areas of business law for clients across multiple industries in over 100 countries worldwide. And while the challenges our clients bring us may vary, we approach and solve them with the same, single-minded focus: finding the smartest way to achieve the biggest impact.

The recent case of HMRC v Gunnarson highlighted the risks of relying on artificial intelligence (AI) for legal advice...
United Kingdom Technology

The recent case of HMRC v Gunnarson highlighted the risks of relying on artificial intelligence (AI) for legal advice, demonstrating that guidance provided by AI is not always accurate or factually correct. Polly Bradley and Hannah Spurr consider the dangers of an overreliance on AI, with a particular focus on personal tax. Court judgments have criticised AI support and stressed that it shouldn't be utilised without appropriate human oversight, and there can be significant financial and legal implications of using AI to cut corners and save money. If you are an individual in need of advice relating to your personal tax and succession planning, the team at Addleshaw Goddard has a wealth of knowledge and would be more than happy to help.

HMRC v Gunnarson

It's safe to say that personal tax is not a favourite hobby for most. None of us want to return home after a long day at work to do more paperwork and look after our personal finances. Unfortunately, the recent case of HMRC v Gunnarson highlights how relying on AI for personal tax and legal advice can backfire. In the case, a taxpayer appealed HMRC's decision to claw back a £13,000 Self-Employed Income Support Scheme (SEISS) grant. The taxpayer was unrepresented, and their case was heavily dependent upon legal arguments drafted by an AI chatbot – which seemed to be convincingly supported by case law.

Unluckily for the taxpayer, the tribunal soon discovered that the sources cited did not exist. The chatbot had 'hallucinated' and generated a number of entirely fictitious cases. Not only was the appeal dismissed, but the individual faced the embarrassment of being criticised by the judge for relying on an AI chatbot to make their case.

A common theme

HMRC v Gunnarson is not the only instance in which fictitious case citations have been submitted to the Court due to AI 'hallucinations', serving as a reminder that AI should be used as a supporting tool, not as a sole adviser. For those not familiar with the complexities and technicalities of personal tax, it becomes difficult to spot flaws in a persuasively drafted AI argument. Not to mention that personal tax rules are constantly changing, increasing the risk of AI citing outdated law and cases. As noted in the judgement of Ayinde, R (On the Application Of) v Qatar National Bank QPSC & Anor) (another recent case which criticised reliance upon AI in legal proceedings), 'Artificial intelligence is a tool that carries with it risks as well as opportunities. Its use must take place therefore with an appropriate degree of oversight'.

Our advisers

HMRC v Gunnarson demonstrates that relying on AI as any kind of adviser for your personal affairs is not a wise decision. For those tempted to cut corners, we would strongly urge that you seek a second, human opinion before relying on any AI generated legal and tax advice. Our Private Capital team here at Addleshaw Goddard have a wealth of knowledge and would be more than happy to help with any queries you may have in relation to your personal tax and succession planning.

The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.

Mondaq uses cookies on this website. By using our website you agree to our use of cookies as set out in our Privacy Policy.

Learn More