Internet Explorer 11 (IE11) is not supported. For the best experience please open using Chrome, Firefox, Safari or MS Edge

How does using AI chatbots impact on legal privilege? Our Dispute Resolution team explores the confidentiality risks of sharing sensitive data with public AI models and discusses how AI-generated drafts might be protected under Irish law.


What you need to know

  • Irish courts have encountered and accepted AI use in certain scenarios but have not ruled on whether privilege applies to AI inputs and outputs.
  • Only qualified human lawyers can provide legally privileged advice under Irish law.
  • Using public AI tools poses severe confidentiality risks, which can destroy legal advice privilege.
  • Using secure AI to draft legal advice may be protected as preparatory work.

Irish courts are already encountering AI in litigation and have indicated that using software to assist with drafting legal correspondence is acceptable[1]. However, the Irish courts have yet to consider the application of privilege over AI inputs and outputs. A recent case in the Southern District of New York[2] found that communications between an individual and a chatbot concerning upcoming criminal proceedings were not privileged, even where the outputs were subsequently shared with an attorney. This case demonstrates that it is only a matter of time before this issue is determined in Ireland. We consider how the Irish courts might treat the application of legal advice privilege in the age of AI.

What is privilege?

Privilege is a legal protection that allows a party to withhold the production of documents in proceedings, provided it meets certain requirements. Parties are required to identify all relevant evidence in proceedings. However, the successful application of privilege allows a party to withhold the disclosure of a privileged document to the opposing party.

This protection serves an important function. It enables a party to openly seek legal advice (legal advice privilege) or prepare for anticipated litigation (litigation privilege). In this article, we will focus on AI and legal advice privilege.

Who is a lawyer?

Privilege in Ireland attaches to communications with lawyers, where the dominant purpose is seeking or obtaining legal advice. The communications must arise in the course of a professional lawyer/client relationship. The individual providing the advice should also be on the roll of solicitors with a valid practising certificate. As an AI chatbot is fundamentally software and not a legal professional, it could not be relied on for maintaining privilege. For example, if a member of a commercial team provided sensitive information to a chatbot for the purpose of seeking legal advice, that would not benefit from privilege as the chatbot is not a lawyer.

Advice versus assistance

Irish courts draw a clear distinction between seeking legal advice, which is privileged, and legal assistance, which is not. Legal assistance can include lawyers providing purely administrative input on a project.

For example, if an in-house counsel asked an AI tool to summarise a commercial timeline, that output is not privileged as it does not amount to the provision of legal advice. However, if an in-house counsel uses AI to help draft legal advice, the prompts and outputs may be protected as the lawyer's work product.

The confidentiality trap

A document or communication must be confidential to attract privilege under Irish law. If a document's confidential character is lost, privilege can be challenged. This is the case even in circumstances where the document clearly contains legal advice.

If a lawyer inputs privileged material into a public AI tool, it is likely that the court will view this as a failure to maintain confidentiality. However, if that lawyer inputted privileged material into a secure AI system with appropriate safeguards, that could be sufficient to maintain confidentiality and claim privilege.

As businesses increasingly rely on AI, establishing clear internal guidelines is vital. If a regulatory investigation or litigation occurs, unprivileged AI inputs and outputs could become discoverable. Businesses need to understand this risk.

Next steps

Fine margins are at play when maintaining privilege. These decisions highlight that integrating AI requires clear policies and strict boundaries within an organisation.

If your organisation would like to learn more about the application of privilege to AI outputs, please contact a member of our Dispute Resolution team.

People also ask

Does legal advice privilege apply to AI chatbot outputs?

Irish courts have not yet decided this issue. However, legal advice privilege only applies to communications with qualified human lawyers. Direct outputs generated by an AI chatbot for a non-lawyer would not qualify.

Can using public AI waive legal privilege?

Privilege requires strict confidentiality. Inputting sensitive legal information or previous legal advice into a public AI model could destroy that confidentiality and waive the privilege entirely.

The content of this article is provided for information purposes only and does not constitute legal or other advice.

[1] Farina v X

[2] U.S. v Heppner (No. 1.25-cr-00503 S.D.N.Y.)



Share this: