Further guidance on the use of AI in legal submissions in Irish legal proceedings

At a glance

• Guidance on the use of AI in legal submissions has been provided by both the Court of Appeal and the Labour Court.
• The guidance highlights the risk of potential AI hallucinations in submissions and emphasises that parties are fully responsible for any submissions lodged to court, even if AI generated.
• Strategies in responding to AI generated submissions may differ depending on the Court venue.

Court of Appeal judgment: James Guerin v Gemma O'Doherty

Ms Justice Costello addressed the use of AI-generated submissions in appeal proceedings. The defendant, a self-represented litigant, had relied on AI to prepare written submissions which cited a number of authorities that did not exist. These hallucinated cases required the plaintiff to incur additional costs attempting to locate them and prompted specific judicial guidance on appropriate use of AI.

Ms Justice Costello articulated a series of principles to be applied regarding AI use including that:

• Parties may use AI responsibly for research purposes.
• Must use AI responsibility and must not, even inadvertently, mislead the court.
• Must disclose its use to the court and other parties.
• Must independently verify the accuracy of submissions and authorities relied on.
• Must not cite any authority that has not been confirmed as genuine and relevant.
• Responsibility for submissions rests with the party using AI, whether represented or not.

The Appeal Court emphasised that the uncritical reliance on AI-generated material leads to wasted time and costs, risks bringing the administration of justice into disrepute and may mislead the court. Importantly, the Appeal Court noted that a range of sanctions may be available in future where AI is used in breach of these principles. This may include awarding costs against the offending party.

Labour Court guidance

The Labour Court has separately issued guidance to assist parties who decide to use AI when preparing written submissions or documentation for proceedings before it. The guidance recognises that AI tools may offer efficiency benefits but highlights the need for awareness of their limits, risks and shortcomings.

The Labour Court guidance identifies various risks associated with AI, including:

• The potential for hallucinated or fictitious information.
• Confidentiality concerns when using public tools.
• The possibility that information may be outdated or not relevant to Ireland.
• The risk of biased or inaccurate outputs.

A central theme of the guidance is responsibility. All parties, whether legally represented or not, remain responsible for the documentation submitted to the Labour Court. Any material prepared using AI should be checked and reviewed, and parties are cautioned against relying on AI for substantive work. Where AI is used, parties may wish to, but are not explicitly required to, reference its use. Parties may be required to explain submissions made to the Court.

Common themes, divergences and practical considerations

Both the Court of Appeal and the Labour Court accept that AI may be used as a research aid and emphasise the need for verification, caution and personal responsibility. Each highlights hallucinations as a central risk and makes clear that parties cannot delegate responsibility for accuracy to AI tools.

However, there are notable areas of divergence. The Court of Appeal requires express disclosure of AI use and expressly references potential sanctions. By contrast, the Labour Court guidance adopts a less prescriptive tone, stating that parties 'may wish to reference its use' rather than stating this 'should' be done. Further, it is not within the Labour Court's remit to award costs – employers defending Labour Court claims may be required in some cases to spend costs defending AI generated submissions containing hallucinations, even in circumstances where the case is won.

Conclusion

These developments signal a clear expectation that AI should support, but not replace, careful legal judgment and verification. Parties and legal representatives should assume that failures to verify or appropriately disclose AI use may attract scrutiny. In an appeal court context, this may extend to sanctions, while in employment proceedings before the Labour Court, improper reliance on AI may undermine the credibility and effectiveness of a party’s case.