FCA authorised · FRN 724952 0117 325 0027Quote & buy →
Insurance wiki
Liability for an AI hallucination usually rests with the business or professional that relies on, publishes or presents the false output, not with the AI tool. It matters because courts are already sanctioning fictitious AI-generated material, and a client's loss from it is usually a professional negligence claim.
Part of: AI insurance UK
In short
An AI hallucination is confident but false AI output, such as an invented case, fact or figure. The AI has no legal personality, so liability falls on people: the firm that used the output in client work, the business that presented it to customers, and sometimes the vendor under contract. UK courts have referred lawyers to regulators over fake AI citations (Ayinde v Haringey, June 2025), and the July 2026 UKJT statement says professionals can be negligent for failing to validate AI outputs. A client's loss is usually a PI claim, subject to the policy terms.
An AI hallucination is output from a generative AI system that is presented confidently but is false: an invented fact, a misquoted source, a non-existent case or a wrong figure.
The High Court put the risk plainly in June 2025. Freely available generative AI tools "are not capable of conducting reliable legal research"; they can give plausible answers that are wrong and cite sources that do not exist (Ayinde, para 6). It becomes a liability question when someone relies on the output.
An AI system has no legal personality, so liability falls on the people and businesses around it. The UK Jurisdiction Taskforce's Legal Statement on Liability for AI Harms, published in July 2026, concluded that existing English common law can deal with AI liability. It is not binding on courts (summary).
| Party | Typical position |
|---|---|
| User firm (a professional using AI in client work) | Owes the usual duty of reasonable skill and care. The UKJT statement says a professional may be negligent for using AI inappropriately, using an unsuitable model or failing to test or validate outputs. |
| Deployer (a business presenting AI to customers) | Responsible for what its AI tells customers. According to reporting of the statement, businesses cannot avoid responsibility simply because a chatbot produced false information. |
| Vendor or developer | Liability to the deployer is usually set by contract, often with accuracy disclaimers and caps. The UKJT considers developers of general-purpose models will not usually be liable for unforeseeable downstream harms. |
False statements about an identifiable person can also lead to defamation claims. On contract allocation, see AI contract clauses and indemnities.
In a judgment of 6 June 2025 the Divisional Court dealt with two referrals. In Ayinde, judicial review grounds cited five cases that do not exist, and an earlier wasted costs order required the barrister and the instructing law centre to pay £2,000 each. In Al-Haroun, 18 of 45 citations did not exist; the client accepted that public AI tools produced them, and his solicitor relied on them unchecked. The court referred the barrister in Ayinde to her regulator and solicitors in both cases to the Solicitors Regulation Authority. Lawyers, it said, have "a professional duty" to check AI research against authoritative sources (para 7).
In a decision of 4 December 2023, a self-represented taxpayer cited nine tribunal decisions. The tribunal found none were genuine and that they had been "generated by an AI system". It accepted she did not know.
A US federal judge imposed a $5,000 sanction under Rule 11 on two lawyers and their firm for submitting non-existent opinions generated by a generative AI chatbot and standing by them when questioned. He noted "there is nothing inherently improper about using a reliable artificial intelligence tool for assistance".
On 14 February 2024 a Canadian tribunal held an airline liable in negligent misrepresentation for its chatbot's wrong answer on bereavement fares. It is not binding in the UK. See AI chatbot insurance.
When a hallucination in AI-assisted work causes a client financial loss, the claim is usually professional negligence, and PI usually responds, subject to the policy terms. What commonly decides cover:
Insurers look for human review, source checks and logs of prompts and outputs. See what to document about your AI use, AI-generated advice and professional negligence and AI insurance UK.
Usually the business or professional that relied on, published or presented the output; the AI tool has no legal personality. A firm passing AI output to a client owes the usual duty of care, and a business is generally responsible for what its chatbot says.
Yes. In Ayinde v Haringey [2025] EWHC 1383 (Admin) the Divisional Court referred a barrister and solicitors to their regulators, after an earlier wasted costs order. Lawyers must check AI research against authoritative sources.
Often, where a hallucination in your professional work causes a client financial loss and you are sued for negligence. It depends on the policy terms, including the business description and any AI exclusion or endorsement.
Possibly, but it is usually hard. Vendor terms commonly disclaim accuracy and cap liability, and the UKJT statement considers developers of general-purpose models will not usually be liable for unforeseeable downstream harms. Your contract with the vendor sets most of your recourse.
Tell us what your AI work is, where you are based, your turnover and the limit you need. It takes a few minutes online, and a broker will come back to you with options. Or call 0117 325 0027.
Start an AI insurance quote Request a callbackApex Insurance Brokers Limited is authorised and regulated by the Financial Conduct Authority. Registered in England and Wales, company number 07014570. This page is general information, not advice on your individual circumstances, and it does not guarantee that cover will be available or on what terms.
