There is no escaping the creep of artificial intelligence (AI) in all aspects of business and society. AI tools are transforming how lawyers operate, enabling faster research, drafting and document analysis. The courts are becoming more familiar with pleadings and submissions founded on some type of AI technology. Earlier this year in the landmark case of Guerin v O'Doherty [2026] IECA 48 the Court of Appeal set down general guidelines for the use of AI in litigation. One area that the courts have yet to consider is the interaction of AI and legal privilege. Given the pace at which the AI landscape is evolving it is likely to be sooner rather than later that the courts will have to rule on this point. In this alert we look at what the courts might consider in relation to AI and privilege. We will also look at what steps clients should now put in place to ensure any legal protection is not lost.
AI is evolving at such a pace that lawyers need to be alive to ongoing risks and threats. The growth of public generative AI tools has created a legal quandary that no lawyer can overlook. Their use engages several tenets of the profession. In particular, one of these is the concept of legal professional privilege (LPP).
LPP protects the confidentiality of lawyer client communications. It comprises both legal advice privilege and litigation privilege. Legal advice privilege protects confidential communication between a lawyer and their client, made for the dominant purpose of giving or receiving legal advice. Litigation privilege protects communication between a lawyer and client or between either of them and a third party (such as expert witnesses or investigators), made for the dominant purpose of conducting existing or reasonably contemplated litigation.
LPP is founded on confidentiality. Privileged communications (written or oral) that pass between a solicitor and their client or a prospective client, the privilege is that of the client and the solicitor cannot be compelled to disclose those communications unless ordered to do so by the courts.
In considering whether privilege applies to AI inputs and output the courts are likely to be influenced by several factors. These include concepts within the traditional framework of privilege that we are already familiar with such as how a "lawyer" is defined and distinguishing between legal advice versus assistance.
Privilege in Ireland attaches to communications with lawyers where the dominant purpose is the seeking or obtaining of legal advice. The communications must be between a client and a qualified legal professional. As an AI chatbot is software, it cannot satisfy this requirement. Arguably if a lawyer sought legal advice from a chatbot and provided sensitive information to it for this purpose this would not be covered by privilege as the chatbot is not a lawyer.
Whether privilege applies also depends on the nature of the information being given. It depends on whether the information is legal advice or legal assistance. Legal advice is privileged whereas legal assistance is not. Legal advice generally involves the exercise of skills and judgement such as analsying/interpreting the law and advising on a strategy or outcome. Legal assistance on the other hand doesn’t involve human judgement. It might include administrative tasks or conveying information. Arguably if a lawyer asks an AI tool to assist with an administrative type task such as providing a timeline or chronology, then this output and its prompt would not be covered by privilege.
Another feature of privilege is that a document must be confidential to attract privilege under Irish law. If a document's confidential nature is lost, then any privilege attaching may also be lost. This is where the importance of using an enterprise system for AI as opposed to a free public platform becomes relevant. If a lawyer inputs privileged material into a public AI tool this may be viewed as a failure to maintain confidentiality and accordingly lose any privilege attached to it. Another potential risk area is the use of AI note takers in client meetings. Having a bot at a meeting allows a third-party access to a conversation. This might be seen by the courts as a waiver of privilege because by sharing the data with a third-party provider it may breach the confidentiality requirement.
Decisions from other common law jurisdictions may also influence judicial thinking. In the case of US v Heppner (S.D.N.Y Feb 2026) a US federal court found that communications between a defendant and a chatbot concerning upcoming criminal proceedings were not privileged, even in the circumstances where the outputs were shared with their lawyer. The court opined that even if they had been privileged that privilege was lost the moment the information was shared with the tool.
In the UK, the court in the Commissioners for HM Revenue and Customs v Marc Gunnarsson [2025] UKUT 00247 (TCC) expressly warns against entering any " information into a public AI chatbot that is not already in the public domain" or information that is "private and confidential" as "any information that you input into a public AI chatbot should be seen as being published to all the world". It also mentions risks that open AI models can mishandle confidential information. Similar reasoning is applied in the case of R (on the application Ayinde) v London Borough of Haringey [2025] EWHC 1383 which highlights the importance of a lawyers' professional duty to check the accuracy of AI generated content.
This is an area to be monitored closely as the use of AI may threaten to undermine a key client protection. AI will not change the framework within which the privilege rules operate. What will change is the document trail. Clients should also be alive to the legal issues surrounding lawyer/client communications when using AI tools so they do not find themselves exposed to unintended consequences from them.
Takeaways
Based on the learnings from other jurisdictions the following guidelines might be applied here as we await a judicial decision in this space:
- Training – understanding how AI tools use your information and how documents/content may be generated. Under the EU Artificial Intelligence Act there is an AI literacy obligation to ensure that employees have a certain level of understanding as to the AI systems being used in their workplace. This training should include an understanding that inputting information in the public domain can breach client confidentiality and result in privilege being waived.
- Guidelines/policy – guidelines on appropriate and acceptable use of AI by employees specifically warning against the use of public AI tools.
- Ensure the correct AI platform is used – preferably enterprise platforms as opposed to free resources where there is a significant risk that privilege will be lost where material is uploaded to a public AI tool. The Law Society of Ireland issued formal regulatory guidelines concerning generative AI in legal practice. One of those guidelines includes an absolute prohibition on consumer tools. It notes that entering any client personal data or sensitive case material into a free or public consumer AI tool is a direct violation of professional confidentiality rules.
This is an area which is evolving rapidly. We will continue to monitor developments and provide further updates.
For more information please contact Thomas O'Dwyer, Sinead Grace or your usual contact in Beauchamps.