The Employment Appeal Tribunal (EAT) provides useful guidance on the ever-increasing use of AI in employment litigation.
Background
Mr Hancox was employed by the National Farmers’ Union (NFU) as a handyperson. He was dismissed in February 2020 during his probationary period, for alleged misconduct.
Mr Hancox then brought a variety of claims (12) before the Employment Tribunal (ET) against both the NFU and four individual managers (including NFU manager, Kenneth Sutherland). The claims included whistleblowing detriment, ordinary and automatic unfair dismissal and disability discrimination, all of which was denied by the NFU and the individual respondents. Mr Hancox acted as a litigant-in-person.
The ET proceedings and decision
A significant feature of the ET proceedings was Mr Hancox’s conduct. For more than a year before the final hearing was due to take place, he began posting on LinkedIn criticising the NFU and the individual managers. This resulted in more than 145 posts about the subject matter of his claims, including documents disclosed to him during litigation. Mr Hancox did not stop posting despite the respondents’ solicitors request and as a result, an application to strike out his claim against the four individual managers was made. Mr Hancox continued posting on LinkedIn even after the application was made.
The ET struck out Mr Hancox’s claims against the four individual managers. Specifically, Judge Lloyd found that Mr Hancox’s conduct during the proceedings had been “scandalous, unreasonable and vexatious” which in turn could affect the four individual managers’ willingness or ability to give evidence i.e., a fair trial was no longer possible.
Unsurprisingly, given Mr Hancox’s conduct up to this point, he appealed the Tribunal’s decision to strike out his claim referring to the unfairness of the trial. He also at a later stage tried (and failed) to argue that there were issues under the European Convention on Human Rights (specifically, Article 6, the right to a fair trial, and Article 10, freedom of expression).
What happened at the EAT
Shortly before the preliminary hearing for the appeal, Mr Hancox, still acting as a litigant-in-person lodged a ‘skeleton argument’ consisting of 300 pages and almost 132,000 words. The relevant practice direction suggests skeleton arguments in the EAT should be between 5-15 pages. He also invited the respondents to identify in writing, before 4pm the following day, any inaccuracies or errors in his skeleton argument. Mr Hancox admitted during the preliminary hearing that the document had been produced using ChatGPT because he did not have enough time.
Mr Hancox’s approach was not well received by the EAT. Marcus Pilgerstorfer KC (sitting as Deputy Judge of the High Court) did not dismiss the appeal and found Mr Hancox’s grounds of appeal regarding procedural fairness reasonably arguable. However, the Judge did not mince his words about the skeleton argument, stating that the word “skeleton” was “entirely inapposite” for that length of document. He then set out what the expectations are for all litigants using AI, whether represented or not. Specifically, parties must take personal responsibility for ensuring that AI-generated material submitted to the Tribunal:
- complies with procedural rules, which is a mandatory requirement, not optional – Judge Pilgerstorfer said “being a litigant in person, or being unaware of the requirements, will not generally provide a good reason for non-compliance”;
- is factually, evidentially and legally (regarding any authorities relied on) accurate so as to not mislead the Tribunal and therefore the onus to ensure this is on the person submitting the document; and
- focuses on the relevant issues so that it is easily comprehensible and does not have unnecessary repetition.
The EAT added that AI use may also make a tribunal more circumspect about the weight it can properly place on a document or evidence.
Comment
We can take away quite a few lessons from the EAT judgement.
- Do not assume that the length of a grievance, appeal or particulars is indicative of merits. There can be a lot of information that obscures the core complaints.
- Conduct during the lifetime of a claim, irrespective of the merits of a claim, can be determinative. Mr Hancox’s claims were never heard because of his conduct during the proceedings.
- Judge Pilgerstorfer’s comments on the ‘personal responsibility’ of litigants for AI-generated materials were made in the context of employment litigation. However, the same reasoning directly applies to workplace matters which precede litigation. HR should consider building clear expectations into grievance and disciplinary policies on the use of AI, including on the format, length and identification of specific complaints or issues. Further, when dealing with an AI-generated document and responding to it, HR should focus on the relevant complaints and not feel that it has to respond to every single point raised.
The decision on Hancox does not warn against using AI (Judge Pilgerstorfer openly recognised the benefit of AI for those “who do not otherwise have access to professional legal assistance”). Rather, it is that AI-generated documents should not be uncritically relied on and the person submitting such a document is responsible for ensuring its accuracy and relevance. As Judge Pilgerstorfer stated, “there is no principled reason why a litigant in person should not take reasonable steps to use [AI] responsibly”.