This blog post warns of the growing dangers of relying on AI for legal advice. Drawing on recent examples from my own practice, I explain how AI can wildly overvalue claims, misrepresent genuine case law, and confidently provide authorities which do not support the propositions for which they are cited. Although AI can be a valuable tool and offer significant productivity benefits, it cannot replace the judgment, experience and specialist knowledge of a solicitor.
Content Authenticity Statement
100% of this week’s blog post was generated by me, the human.
AI Legal Advice: Why Artificial Intelligence Is Creating Problems
The Economist magazine recently reported (“How AI is breaking the British State”, 6 August 2026) how the tide of AI generated content is beginning to cause problems for legal and governmental systems built for the “age of the post office and the telephone box”. The backlog in Employment Tribunal claims has risen by 55% in a year, as a result of “AI- fuelled claims”. There is a warning here for us all, as many of those claims will unfortunately have no merit, but the AI programme which a person has turned to for “advice”, such as Claude, ChatGPT or Co-Pilot, probably won’t tell them that, as it quite literally has no “skin in the game” and has a tendency to flatter and encourage its users – promising them that what they want is achievable, even when it isn’t.
When AI Gets Compensation Claims Wrong
By way of illustration of this, I can share some very recent experiences of my own. The first concerned a client who has a good claim against the Police for a wrongful “stop and search” incident – but who was led by AI research to wildly over-value his claim.
After medical evidence had been obtained in order to assess the psychological impact of the incident, my client informed me that his realistic estimation for total compensation was in the region of £150,000. Unfortunately, this was out of all proportion to his actual injuries and losses, and was clearly the result of his making an enquiry with an AI programme.
I, and two of my colleagues, who between us have around 100 years of litigation experience in the fields of personal injury and actions against the Police, in fact valued this client’s claim as being worth in the region of £10,000 – £15,000 i.e an entire order of magnitude less than the AI advice.
AI Cannot Replace an Experienced Actions Against the Police Solicitor
“Large Language Models” (AI) are not a substitute for the experience and depth of knowledge of a solicitor such as myself who has practiced in this field of law for over 30 years. AI programs are notorious for misunderstanding, misquoting or indeed completely hallucinating the law and are designed to ‘please’ their users by telling them what they want to hear – even if what they are saying has no basis in reality.
AI Hallucinations, Fake Case Law and the Risks of AI Legal Research
Indeed, there have been well- reported cases of AI programmes feeding their users entirely made-up case reports, and advocates who don’t know better getting themselves into a great deal of trouble by seeking to rely upon these “Frankenstein” cases before a human (– well, probably-) Judge.
When AI Misinterprets Genuine Case Law: AXX v Zajac
Another of my clients referred me recently to two cases which AI had told him supported his contention that a vulnerable Claimant with psychiatric problems should be able to refuse examination by the Defendant’s medical expert on grounds of “re- traumatisation”, and could instead have the Court only rely upon the Claimant’s own, already obtained report, contrary to the age-old “equality of arms” presumption in litigation which is that both parties should be allowed to call upon equal amounts of expert evidence.
As it happened, these cases really did exist – but they really did not support the argument which AI was telling my client they did. One in particular was the case of AXX v Zajac [2022] EWHC 2463. Whilst AXX was indeed a vulnerable Claimant, who at one stage had been sectioned under the Mental Health Act, whose identity was anonymised for the purposes of the litigation, and who was suspected to have suffered a traumatic brain injury, the Judge in the case highlighted the importance of Claimants “cooperating with and speaking to medical experts for the purpose of expert reports for the Court” and went on to make an order that AXX had to make himself available for interview and assessment by the Defendant’s experts, and that if he did not and instead continued to refuse to engage with the Defendant’s medics then a “Stay or De-barring Order” could be imposed, effectively preventing the claim from proceeding.
So, the conclusion of that case was really the complete opposite of what the ‘cheerleading’ AI bot had led my client to believe it was.
The Danger of AI Telling Legal Clients What They Want to Hear
The root cause of this problem is the AI’s desire to please its user by providing them with material to support their argument, even when, in reality, that material simply does not exist. AI programmes are not people and have no sense of shame or morality; point out to them their error and they are simply likely to respond with an electronic shrug and Cheshire- Cat smile-
“Good catch! You’re right to question that, I made it up – my apologies.”
Cold comfort for those who have thereby been misled in high-stakes civil or criminal litigation cases.
AI Is a Useful Legal Tool – But It’s Not a Lawyer
AI is a very useful tool, and can have real productivity benefits – but we mustn’t lose sight of the fact that it is a tool; effectively, it is a very sophisticated search-engine, it is not a true author or adviser as it does not understand, or indeed care about, the highly polished articles it produces and presents to you. Human intelligence in any field of expertise must remain the real deal and the last resort. I hope that this warning will help to prevent people with good cases self- sabotaging them by reliance upon swift, smooth, polished, but ultimately unreliable AI.
Also remember that you can generally spot an AI article, essay or letter by its almost invariable habit of adopting the following structure for its arguments: “[Something] is not X. It is Y” e.g “The future of legal isn’t fewer lawyers. It’s fewer decisions requiring lawyers.”
Watch out for such language, and superficially persuasive information which won’t stand the stress- test of reality. If you don’t, you may be in danger of TerminAIting your case, or as AI-Arnie might say:
“I don’t need your clothes, your boots, or your motorcycle.
I need all three of them.”
How you can help
I publish these articles regularly to demystify police powers and to help people understand when those powers have been misused. If this post has clarified your rights or reinforced the importance of police accountability, please consider leaving a 5-star review. Your review helps point others towards experienced representation at a time when clear guidance really matters. Thank you.
Contact the Expert Police Misconduct Solicitor
Iain Gould is a solicitor specialising in complaints, claims and civil actions against the Police. With over 30 years of experience and a national reputation, he has successfully sued all 43 police forces in England and Wales challenging abuse of power and securing rightful compensation
You must be logged in to post a comment.