The ChatGPT Legal Trap: Flawed AI Submissions Hit UK Tribunals

The Ghost in the Machine

Employment Tribunals were conceived with a noble, almost rustic, vision: a forum for swift, simple justice where workplace disputes could be resolved through common sense, far from the arcane complexities and prohibitive costs of the traditional courts. The initial idea was that lawyers would be the exception, not the rule. Fifty years on, that vision has become a distant memory. Tribunals are now a highly legalistic battleground, yet the system is still fundamentally designed to accommodate individuals representing themselves – the so-called ‘litigants in person’. Now, a new and disruptive force has entered this already strained arena: generative artificial intelligence. For many unrepresented claimants, AI tools like ChatGPT appear to be a miracle, a digital advocate capable of drafting professional, articulate, and forceful legal arguments. However, legal professionals on the receiving end are discovering a troubling reality. The ‘robot lawyer’ has arrived, but it is one with immense confidence, a poor grasp of the law, and a worrying tendency to invent the facts.

Algorithmic Advocacy: A New Challenge for Employers

For law firms representing employers, the rise of AI-generated submissions has become a significant and growing frustration. A distinct pattern is emerging in the grievances and tribunal claims landing on their desks. These documents are often polished and articulate, but they possess an uncanny uniformity. Legal experts note that arguments frequently appear in neat clusters of three, are littered with corporate buzzwords, and convey a palpable sense of outraged dignity that feels formulaic rather than genuine. The language is sophisticated, but it lacks the specific, nuanced detail of a personal account. It is advocacy by algorithm, and it presents a unique challenge. While the document may look impressive, it often betrays a complete absence of genuine legal or procedural understanding. The AI, acting as a confident but incompetent adviser, is creating a mountain of work for respondents, their legal teams, and ultimately, the tribunals themselves, who must patiently unpick these well-written but often nonsensical submissions.

The Perils of AI-Generated Justice

The promise of democratising legal representation through AI is seductive, but the reality is fraught with danger for the very individuals it purports to help. The issues go far beyond mere stylistic quirks, striking at the heart of legal accuracy and personal credibility.

Phantom Precedents and Fabricated Law

One of the most alarming traits of current generative AI is its capacity for ‘hallucination’ – inventing information with absolute conviction. In a legal context, this is disastrous. AI-drafted submissions have been found to contain references to phantom legal cases, citing precedents that simply do not exist or misrepresenting the outcomes of real ones. Whereas a human-drafted grievance typically focuses on the factual narrative of events, AI tends to pepper its output with more legal citations than a seasoned barrister might use. For an employer or a tribunal, verifying these claims consumes valuable time. For the claimant, building a case on a foundation of non-existent law is a strategy doomed to fail, potentially collapsing their entire argument under scrutiny.

A Detachment from Reality

When an employee describes a workplace incident they have personally experienced, their account carries the weight of authenticity. When that same incident is described by an AI, which has no experience, emotion, or true understanding, the narrative can become exaggerated or strangely detached from reality. The AI optimises for persuasive language, which can twist a simple disagreement into a dramatic confrontation or a minor oversight into a calculated conspiracy. This not only presents a distorted picture to the employer but also poses a serious risk to the claimant’s credibility. If they are unable to substantiate the AI’s dramatic retelling of events when questioned in a tribunal, their testimony can be undermined entirely.

Unrealistic Expectations and Data Privacy Nightmares

Large language models are trained to be helpful and provide positive, solution-oriented responses. This programming creates a significant optimism bias. An employee using AI may receive overly confident advice on the merits of their case, leading to unrealistic expectations about potential outcomes or compensation. This can make settlement negotiations impossible and prolong the dispute unnecessarily. Furthermore, there is a critical and often overlooked issue of data privacy. When an employee inputs the sensitive, confidential details of their employment dispute into a public AI model, that information is no longer private. It can be absorbed into the model’s training data, effectively releasing personal and company secrets onto a public network, with profound implications for privacy and data protection.

A Warning from the Irish Workplace Relations Commission

While the UK grapples with this emerging issue, a pragmatic approach has already been taken across the Irish Sea. A recent case before Ireland’s Workplace Relations Commission (WRC) directly confronted the use of AI in a complainant’s submissions. The Adjudicating Officer, in a sensible and measured decision, did not prohibit the use of AI as a tool. However, noting the significant errors and misleading information that had resulted from its use, the Officer issued a stern and unequivocal warning: “Parties are reminded that their submissions must be relevant and accurate and do not set out to mislead either the other party or the Adjudication Officer.” Following this, the WRC has formalised its guidance, cautioning that AI-generated content can impede cases, undermine a complainant’s credibility, and cause significant delays. The WRC now explicitly advises employees to double-check all legal references, to fully understand and be able to explain every part of their submission, to avoid using confidential information in public AI tools, and crucially, not to rely on AI for legal advice or case strategy. This guidance serves as a valuable reminder that AI can be a word processor, but it cannot be a substitute for personal understanding and responsibility.

A Call for Clarity in the UK

The Irish example provides a clear and practical blueprint that many believe the UK should follow. The influx of flawed, AI-generated cases is adding a significant burden to an already overloaded Employment Tribunal system. There is a growing call for official guidance from the Presidential Body of the Employment Tribunals, and for practical advice to be featured on the ACAS website. While artificial intelligence is not always wrong, its current application in workplace dispute resolution is proving to be deeply unhelpful. Without clear guardrails, the risk is that the pursuit of simple, common-sense justice will be further obstructed by the eloquent, confident, and dangerously misleading voice of a machine.