EAT criticises disproportionate use of AI in proceedings

Background

Paragraph 11.6.6 of the EAT Practice Direction provides, among other things, that skeleton arguments submitted to the EAT should be as concise as possible, and generally between 5 and 15 pages long. Non-compliant skeleton arguments may be rejected by the EAT and other sanctions imposed. Being a litigant in person, or being unaware of the requirements, will not generally provide a good reason for non-compliance.

Facts

Mr Hancox, a litigant in person, had his claim against four individual respondents struck out by the employment tribunal, which found that he had conducted the proceedings in a scandalous, unreasonable, and vexatious manner through a social media campaign on LinkedIn, such that a fair trial was no longer possible.

He appealed to the EAT, and in support of his appeal filed a skeleton argument created using ChatGPT. The skeleton was 300 pages long and contained almost 132,000 words. It did not comply with the EAT Practice Direction and was not ultimately relied upon at the preliminary hearing, where, with the assistance of counsel, Mr Hancox advanced a single ground of appeal.

Decision

The EAT held that the single ground of appeal pursued at the preliminary hearing was reasonably arguable and should proceed to a full hearing. However, it emphasised that all litigants before the EAT, whether represented or not, must comply with the EAT Rules and Practice Direction, including the procedural requirements for skeleton arguments.

Taking the opportunity presented by the case, the EAT gave guidance that is plainly intended to equip tribunals to deal with the disproportionate use of generative AI. Its principal points are:

• Responsible and ethical use of generative AI may be of benefit, especially to litigants in person, but the use of generative AI also involves risks, including that the output is unfocused, excessive, or misleading, is affected by biases, or contains hallucinations.
• The risks of generative AI are now widely known, and there is no reason why a litigant in person should not use it responsibly. The underlying principles of personal responsibility and accuracy apply to all persons who submit documents to a court or tribunal.
• Litigants who use AI must ensure that all documents submitted comply with applicable procedural rules. In the employment tribunal, this means complying with the overriding objective, which requires litigants to assist the tribunal to deal with the issues in a proportionate manner.
• Litigants must check the documents as thoroughly as they are reasonably able for accuracy, so that factual, evidential and legal points, including references to authorities, are correct and the tribunal is not misled.
• Documents must contain only relevant points, focusing on the central or best arguments, presented comprehensibly and avoiding undue repetition.
• Simply submitting the product of generative AI, or placing the onus to check a document onto an opponent, is not acceptable.
• Where concerns arise that necessary checks have not been undertaken or were inadequate, judicial enquiries and potential sanctions are likely to ensue.

The clear message from the EAT is that, while AI is not generally prohibited for tasks such as preparing a skeleton argument, all litigants must take personal responsibility for any document submitted and it must be carefully checked, as thoroughly as the author is reasonably able.

What does this mean for employers?

This case is an early sign that the EAT (and by extension employment tribunals) will take steps to prevent litigants from using generative AI to make unnecessarily lengthy submissions. This provides employers with cautious optimism that the time and costs spent defending AI-driven claims may be mitigated by robust case management.

Where an opponent serves disproportionate material that bears the hallmarks of AI, employers and their representatives may consider inviting the tribunal to apply the EAT's guidance in this case, rather than shouldering the burden of responding to every point. The guidance is clear that the onus of checking an AI-generated document does not fall on the receiving party.

The same standards apply to employers and their representatives. While there may be ways to reduce the time and costs involved in defending litigation by using AI, documents prepared with AI assistance must be checked for accuracy and relevance and must comply with the applicable rules.

Hancox v Sutherland & Others [2026] EAT 139