TLDR: A recent UK trademark appeal highlighted significant issues with ‘AI hallucinations’ in legal documents, as both the self-represented plaintiff and the defendant’s lawyer presented briefs containing erroneous or unsupported citations, strongly suggesting the use of generative AI. The judge noted a growing concern over the reliability of AI-generated legal content, emphasizing the need for lawyers to verify information and maintain professional standards.
In a recent UK trademark appeal, the legal community witnessed a concerning display of ‘AI hallucinations’ within submitted briefings, impacting both the self-represented plaintiff and the defendant’s legal counsel. The case, detailed by Eugene Volokh on Reason.com’s The Volokh Conspiracy on July 15, 2025, underscores the emerging challenges of integrating generative artificial intelligence into legal practice.
The plaintiff, representing themselves, openly admitted to drafting their documents with the assistance of ChatGPT. The presiding judge noted ‘numerous errors in the citations and problems with the skeleton’ argument produced by the AI. Furthermore, the AI’s output included ‘fabricated citations’ and presented factual issues that were ‘largely not relevant’ to the case, rendering the proposed arguments unhelpful.
Even more notably, the defendant’s lawyer, Mr. Caddy, was strongly suspected of similar reliance on generative AI. His skeleton argument, dated June 6, 2025, cited three existing cases, but the judge found it ‘unclear… the cases cited stand for the propositions claimed by Mr. Caddy.’ During the hearing, Mr. Caddy was unable to identify the specific paragraphs in the judgments that supported his propositions, stating, ‘I cannot actually remember that now, to be honest with you.’ He later vaguely attributed the references to ‘a previous edition of Kerly’s Law on Trade Marks’ or ‘Wadlow,’ neither of which, upon the judge’s verification, contained the asserted propositions.
The judge expressed ‘great concern’ over three aspects of Mr. Caddy’s conduct: his inability to recall the source of legal propositions in a document less than a week old, his apparent belief that he was not expected to elaborate on points in his skeleton argument, and his use of outdated textbooks. The court emphasized the critical need for all lawyers to access ‘relevant and up-to-date textbooks and relevant case reports.’
While the judge ultimately decided against referring Mr. Caddy to the Intellectual Property Regulation Board (IPREG) on this occasion, a stern warning was issued. Mr. Caddy was advised to ‘seriously reflect on how he conducts his practice in the future, including how he undertakes legal research, how he drafts skeleton arguments, and he must ensure that when he appears before a tribunal he is properly prepared to do so.’ The ruling clarified that a ‘fabricated citation’ extends beyond entirely made-up cases to include citing a case for a proposition for which there is ‘absolutely no basis.’
Also Read:
- Legal AI Forum Gathers Industry Leaders to Forge Generative AI Community in Law
- Student Concerns Mount Over AI’s Impact on Learning and Trust in Universities
This incident highlights a growing trend, as evidenced by Damien Charlotin’s AI Hallucination Cases database, which now lists 210 cases (over half from the U.S.), and Peter Henderson’s AI Law Tracker, with 223 cases. The case serves as a stark reminder of the imperative for human oversight and rigorous verification when utilizing artificial intelligence in professional legal contexts.
