The courts are finally taking a stance on the use of generative AI in litigation. In Hancox v Kenneth Sutherland and others [2026] EAT 139, the Employment Appeal Tribunal (EAT) addressed the use of generative AI in documents filed by a litigant in person. This marks the first time a UK Tribunal has taken a firm stance against the use of AI in litigation.
The EAT stressed that litigants who use AI must be personally responsible for checking what they file.
Background
Mr Hancox (H) claimed automatic unfair dismissal, unfair dismissal, whistleblowing detriment and disability discrimination against four individuals (Respondents). The Tribunal struck out his claim for unreasonable conduct and H subsequently appealed the decision.
On appeal and just two days before the hearing, H submitted a 300-page skeleton argument generated entirely by ChatGPT with over 132,000 words. H accepted that the document lacked credibility and the judge would be unable to read his argument so instead, he invited the Respondents to identify any errors in his argument by 4pm the next day.
The EAT held that the filing of unchecked AI-generated material and the document itself were entirely unacceptable. The rules state that a skeleton argument must be between 5-15 pages. This judgment confirms that documents of this kind will now usually be rejected.
Importance
The use of generative AI in employment litigation has been growing over the past few years. Solicitors, barristers and tribunal judges are more commonly finding themselves flooded with excessive pages of AI-generated material.
Whilst AI can, in some ways, make it more accessible for individuals to access justice, there are significant risks attached to its use. Generative AI can often be misleading, false or produce hallucinations of caselaw and facts that do not exist.
The submission of unchecked AI-drafted documents adds further strain and delay on the justice system. Processes are flooded with points that are irrelevant to the facts which distract from the real issues. Going forward, misusing AI in litigation could put litigants at risk of sanctions or having their claims struck out.
EAT Guidance
The EAT set out that anyone filing AI-generated documents must ensure, as a minimum, that:
the document complies with the procedural rules;
it is checked for accuracy as thoroughly as the author reasonably can manage;
factual, evidential and legal points (such as case law refences) are correct and that the tribunal is not misled; and
only the relevant points are included without undue repetition.
It was also confirmed that the responsibility for checking the accuracy of AI generated content lies with the party submitting it – it is not acceptable to place the onus on the opposing party as H did to the Respondents.
Steps for Employers
Employers should take extra steps to challenge submissions of opposing parties where they become aware of or suspect that AI-generated content is being used without proper checks.
We recommend employers first confirm if AI has been used and question whether those documents have been properly reviewed before submission. If unchecked AI content continues to be submitted to the Tribunal, and the Tribunal have not yet flagged such issue, employers should raise this with the Tribunal.
Getting on top of these issues early can potentially avoid lengthy disruptions and heavy workloads reviewing lengthy, AI generated documents. In turn these can save costs, time and needless strain on the justice system.
BPE’s Employment team offer legal advice to both businesses and individuals. If you are looking for legal advice relating to the contents of this article, please get in touch with one of the team who will be happy to help you.
