Key Contacts: Damien Young – Partner | Eimear Collins – Partner| Gerald Byrne – Partner |
As Irish courts begin to address AI-assisted material in proceedings, solicitors remain responsible for ensuring that legal work is accurate, properly grounded and capable of being stood over.
Introduction
AI is increasingly finding its way into litigation practice. Its use is developing against a clear professional backdrop: Irish solicitors remain responsible for the accuracy, confidentiality and reliability of the work they produce, and Irish courts are beginning to address the consequences of AI-assisted material appearing in proceedings.
As AI becomes more capable of producing fluent legal content, the role of professional judgment in checking, contextualising and standing over that content becomes more important. For litigators, the central issue is how AI-assisted work is verified, reviewed and brought within the professional responsibility of the lawyer relying on it.
AI Before the Irish Courts
Irish courts are beginning to encounter AI-related issues in proceedings more and more.
In Reddan v An Bord Pleanála [2025] IEHC 172, the High Court refused an application for leave to bring judicial review proceedings concerning a planning decision. The applicant had relied on language which the Court considered unfamiliar in Irish jurisprudence and which appeared to have been derived from online or AI-assisted research. The case illustrates a practical litigation issue. AI-assisted material may introduce language, concepts or propositions which appear legalistic while sitting uneasily within Irish law. Where that occurs, the argument may lose force because it has not been anchored in the applicable legal framework.
In Von Geitz v Kelly & Ors; Von Geitz v Robertson & Ors [2026] IECA 29, the Court of Appeal made clear that a litigant who uses a tool to generate legal material remains responsible for checking what it produces before serving it on opponents or placing it before the Court. The decision sits alongside Reddan as an example of the courts’ concern with AI-assisted material that has not been properly verified, particularly where it causes unnecessary work for other parties or distracts from the legal issues actually before the Court.
In Guerin v O’Doherty [2026] IECA 48, the Court of Appeal addressed the use of AI in the preparation of submissions. The decision concerned AI-generated material which included authorities that did not exist or did not support the propositions advanced. The Court emphasised the obligation on parties to avoid misleading the court and the importance of verifying authorities and legal propositions before relying on them.
In Farina v X Internet Unlimited Company [2025] IEHC 514, the High Court considered, among other matters, a complaint that legal correspondence from the defendant appeared to have been generated by a computer system, perhaps using AI. Barrett J declined to criticise the use of AI in that context and indicated that there was no legal requirement for such correspondence to be generated by a human. That decision forms an important part of the Irish picture. It suggests that the courts are likely to look closely at context. The use of AI in legal correspondence may be treated differently from reliance on AI-generated legal propositions or authorities in proceedings.
Beyond Fabricated Authorities
The most visible AI failures in litigation involve fabricated cases, and there are plenty of publicised examples of those in many jurisdiction world-wide. Those errors are serious and, once checked, can usually be identified quickly. A more subtle difficulty arises where the authority exists, but the AI output misdescribes it. That may include an inaccurate summary of the facts, an overstatement of the ratio, an omission of a limiting factor, or the use of a case for a proposition which it does not support.
This is of fundamental importance in litigation because legal analysis often depends on fine distinctions. A case may be persuasive rather than binding. A principle may apply in one procedural context and require caution in another. A passage may be obiter dictum rather than ratio decidendi.[1] A decision from another common law jurisdiction may be useful where its statutory or procedural setting is properly understood.
Recent academic work has made this point in the context of legal interpretation, noting that large language models may generate legal outputs which appear persuasive while lacking the reliability and legitimacy required for legal interpretation.[2] The concern extends beyond invented material to the quality of the interpretive process itself.
For litigators, this is where professional judgment becomes particularly important. A fabricated authority is usually exposed through verification. A misapplied authority requires the lawyer to read the judgment, identify the relevant principle, assess its status and decide whether it properly supports the argument being made.
Legal Judgment in AI-Assisted Litigation
Litigation is a methodical process of analysis and judgment. Authorities must be chosen for a reason. Principles must be stated accurately. Facts must be connected to the correct legal test. Procedural context matters. AI can assist with parts of that work. It may be useful in producing working summaries, organising issues and generating first drafts. Used in that way, it can support legal work, provided its output is reviewed by a lawyer with the necessary expertise and experience to assess it properly.
The broader conversation around AI and litigation services is also beginning to evolve. In recent proceedings in the United States, Nippon Life Insurance Company issued federal proceedings against OpenAI, alleging unlicensed legal practice arising from a claimant’s use of ChatGPT to prepare post-settlement filings, one of which included a hallucinated case citation.[3] OpenAI’s position was that ChatGPT is a tool rather than a lawyer and should be treated as a research and drafting aid, not as a replacement for professional advice. The case raises a related question as to whether, and to what extent, responsibility may shift towards the developer of the tool. How that argument develops remains to be seen.
The difficulty arises when fluent AI-generated language is taken as a proxy for legal analysis. A polished paragraph may still contain an inaccurate legal proposition. A confident summary may miss the feature of a case that limits its application. A neat formulation may draw on language from another jurisdiction without recognising the difference in procedural or statutory context.
This is particularly relevant in Ireland, where litigation can involve a careful mix of Irish authority, persuasive common law authority, EU law, constitutional principles and detailed procedural rules. AI-assisted work must be assessed against that framework. It is the lawyer’s task to decide what is relevant, what is reliable and what can properly be advanced.
What this means for Irish litigators
AI will continue to become a familiar feature of litigation practice in Ireland. Its value will depend on how it is integrated into day-to-day litigation work and the care applied to its outputs. The use of AI in litigation raises a number of connected issues. This article focuses on the professional judgment required when AI-assisted material is used in proceedings. Questions concerning legal professional privilege and the broader impact of AI on legal practice are distinct issues which warrant separate treatment.
The early Irish authorities point in a clear direction. AI can assist with legal work, but responsibility for the final product remains with the lawyer who relies on it. That responsibility includes checking authorities, testing the propositions advanced and ensuring that the reasoning placed before the Court is accurate and properly grounded.
For solicitors, the central point is practical. Where AI is used, the final work product must still bear the mark of professional judgment: properly sourced, properly reasoned and capable of being stood over in court.
This article was co-authored by Knowledge Lawyer, Melanie Ardiff.
[1] The ratio decidendi of a case is the legal reason for the court’s decision. It is the part of the judgment that decides the case and may bind later courts in similar cases. Obiter dictum refers to comments or observations made by a judge which are not necessary to decide the case. Obiter comments are not binding, although they may be persuasive.
[2] See James Grimmelmann, Benjamin L.W. Sobel and David Stein, “Generative Misinterpretation” (2026) 63.1 Harvard Journal on Legislation; see also Matthew Dahl, Varun Magesh, Mirac Suzgun and Daniel E. Ho, “Large Legal Fictions: Profiling Legal Hallucinations in Large Language Models” (2024) 16 Journal of Legal Analysis 64.
[3] Nippon Life Insurance Company of America v. OpenAI Foundation and OpenAI Group PBC, No. 1:26-cv-02448 (N.D. Ill.).
