Ad/Marketing Communication
This legal article/report forms part of my ongoing legal commentary on the use of artificial intelligence within the justice system. It supports my work in teaching, lecturing, and writing about AI and the law and is published to promote my practice. Not legal advice. Not Direct/Public Access. All instructions via clerks at Doughty Street Chambers. This legal article concerns AI Law.
Introduction
This post explores the recent Employment Tribunal case Lwanga v AtkinsRealis UK Ltd (Case Number: 2202508/2023), heard from 9-14 February 2024 before Employment Judge Baty. It focuses on how AI, specifically ChatGPT, was used by the Claimant in drafting parts of his submissions (“the AI Submissions”) and the ramifications this had on the Tribunal’s findings and outcome.
Case Background and What It Was About
This case involved a claim of direct race discrimination brought by an individual (the Claimant) against AtkinsRealis UK Ltd (the Respondent). The Claimant alleged multiple acts of discrimination, largely related to his experience working on a specific Heathrow Sprint Project and his subsequent dismissal. The Tribunal heard extensive evidence about how the Claimant was managed, the standards of his work, and various interpersonal issues.
Ultimately, the Tribunal dismissed all allegations of race discrimination. It noted that the Respondent had valid reasons for its actions, such as concerns over the quality of the Claimant’s work, complaints from the Heathrow client, and the Claimant’s interpersonal behaviour. A notable feature was that while the Claimant’s grounds for claiming discrimination centred on his being treated differently due to race, the Tribunal found insufficient evidence to support this. The Claimant’s probation period was first extended and later terminated purely because of performance and conduct issues rather than anything related to race.
Additionally, the Claimant was ordered to pay a costs award of £12,500. This is a considerable sum and highlights that if a claim is deemed to have no reasonable prospect of success and is conducted unreasonably, an Employment Tribunal can order the claimant to contribute to the legal costs of the respondent.
How AI Featured in the Case
An intriguing aspect was the Claimant’s admission that he used ChatGPT to help draft his written submissions. Although generative AI can rapidly produce large amounts of text, it appeared that certain sections of the Claimant’s documentation were inconsistent, replete with grammatical errors, and at odds with the Tribunal’s factual findings. The Tribunal noted the stark difference between passages seemingly produced by AI and the Claimant’s own insertions, which were often riddled with mistakes or cut-and-paste errors.
The Tribunal’s view underlined that while AI tools can be used to streamline legal or administrative writing, the author remains responsible for verifying correctness and clarity. In this situation, the Tribunal found the submissions unpersuasive and not properly tailored to the specific allegations, which were themselves found to be lacking. The case demonstrates the importance of ensuring that any AI-driven submissions still comply with evidential and procedural standards required in legal proceedings.
Comment
This judgment underscores the importance of not over-relying on AI for legal submissions, particularly when no diligent human review is in place. Although technology can accelerate drafting, it can also introduce inconsistencies and omissions if its output is not meticulously checked. Employment Judge Baty emphasised this point as follows:
“51. The claimant then said to the tribunal that he would use ‘Chat GPT’ to produce his submissions and would ‘see what came out’. The judge said that, in his knowledge of other occasions when unrepresented parties had sought to write their submissions using Chat GPT, this had resulted in submissions which were not necessarily tailored to the issues and were of little help to the tribunal. However, the claimant nonetheless indicated that that was what he was intending to do.”
In light of such remarks, when a judge specifically signals that a chosen approach is of limited value to the Tribunal, prudence dictates adopting a more careful strategy.
Indeed, AI-generated content often lacks the nuance and precision required in legal matters. Errors can undermine the coherence of a claim, ultimately harming the party’s credibility. It would be a serious injustice if a deserving claimant were to lose out entirely because of a poorly handled AI submission.
At the same time, the potential of AI in legal drafting is too significant to overlook. In fact, lawyers need to think carefully about whether the refusal to utilise AI in legal practice will eventually be viewed as professional negligence?
If AI is used correctly, it could benefit courts, lawyers, and litigants, harnessing speed and flexibility while preserving thoroughness. The key is to combine AI’s efficiency with sharp-eyed human oversight and legal expertise. I often refer to this as a synergistic partnership between the lawyer and an LLM. This ensure that submissions are factually accurate, tailored to the issues, and free from misleading errors.
