Source: DLA Piper Knowledge

In Ireland on 26 May 2026, the Court of Appeal considered AI-generated written submissions in James Guerin v Gemma O'Doherty. A self-represented defendant had relied on AI that produced authorities which did not exist, forcing the plaintiff to incur costs trying to locate them. Ms Justice Costello set out principles permitting AI as a research aid but requiring parties to verify outputs and not to cite unconfirmed authorities. The court required disclosure of AI use to the court and other parties and emphasised that responsibility for the accuracy of submissions rests with the submitting party. The Appeal Court warned that uncritical reliance on AI wastes time, risks misleading the court and may bring the administration of justice into disrepute. It noted that a range of sanctions may be available in future for breaches of these principles. Those sanctions may include the awarding of costs against a party who uses AI inappropriately. The court framed AI as supportive of, but not a substitute for, careful legal judgment and verification.

Separately, on the same date the Labour Court issued guidance recognising that AI tools may offer efficiency benefits while highlighting limits, risks and shortcomings. The Labour Court identified risks including hallucinated or fictitious information, confidentiality concerns when using public tools, information that may be outdated or not relevant to Ireland, and biased or inaccurate outputs. It emphasised that all parties, represented or not, remain responsible for documentation submitted and should check and review any material prepared using AI. The Labour Court guidance states parties "may wish to reference" AI use rather than mandating disclosure and, unlike the Appeal Court, it cannot award costs. The guidance warned that improper reliance on AI may undermine a party’s credibility and effectiveness, and parties may be required to explain AI-assisted submissions to the Court. Both courts agreed that verification, caution and personal responsibility are essential and that parties cannot delegate accuracy to AI tools; failures to verify or disclose may attract scrutiny and, in appellate contexts, potential sanctions.