TL;DR
Marcus Pilgerstorfer KC, sitting in the EAT, has told litigants in person that they, like lawyers, are responsible for checking anything AI helped them write. The warning followed a 300-page, 132,000-word skeleton argument produced with ChatGPT. Similar filings in future are likely to be rejected outright.
The case
The judge was deciding whether Terence Hancox had reasonable grounds to appeal the 2021 striking out of his claim. Hancox filed the ChatGPT-drafted skeleton and gave the respondents until 4pm the following afternoon to flag any errors.
The judge rejected that approach. “Simply submitting the product of generative AI to a court or tribunal, or placing the onus to check a document onto an opponent, is not acceptable,” he said.
He did not ban AI. He said AI use “is not generally prohibited” for drafting skeletons and similar work, and noted that the EAT is seeing it used for notices of appeal and correspondence as well. But the personal responsibility principles from the 2025 Ayinde judgment, which was mainly concerned with lawyers’ professional duties, extend to “all persons who submit documents to a court or tribunal”. In practice, anyone relying on AI must check the document as carefully as they reasonably can, make sure authorities and facts are correct, and keep to the central arguments without repetition.
Hancox was allowed to proceed on one ground, with case management directions meant to stop it absorbing more than its share of tribunal resources.
Why this matters now
The ruling follows JUSTICE research, reported this week, finding that one in six people in a legal dispute now use an AI chatbot. Some of those will be representing themselves, and this judgment tells them the tribunal will not do their fact-checking.
The American experience suggests sanctions alone do not solve the problem. Reuters reports that AI hallucinations have now turned up in 1,395 or more US state and federal cases, some three years on from the first sanction over fake ChatGPT citations. Across a three-day span in September, a California lawyer acting for State Farm was fined, an Oklahoma judge’s order was found to cite non-existent cases, and a New Mexico defence lawyer was rebuked over invented police testimony. Judge P. Kevin Castel, who issued the 2023 sanction, told Reuters that tougher deterrents are needed as the problem keeps growing.
Looking forward
The EAT has set a clear expectation that procedural rules apply regardless of who or what drafted the document. The harder question is practical: litigants in person often lack the skills to verify legal citations at all. Expect pressure on HMCTS to publish plain-language guidance, and possibly a rule change requiring parties to declare AI use.