AI and legal privilege – Do you know your risks?

In the legal community around the world we have heard of AI hallucinations, where someone asks AI for help and it provides a made up case authority. This troubling trend has caused negative comments from many courts and tribunals and has led to guidance as outlined here.

Now thanks to an immigration case from November 2025, we have a UK tribunal telling us that there is also a risk of losing legal professional privilege if you use AI. This is a very serious concern and needs to be understood.

What is legal professional privilege?

Legal professional privilege is an essential safeguard in the legal system. Lawyers talk of privilege as an absolute right, which belongs to the client. It exists as a protection to the individual to entitle you to tell your lawyer anything without anyone else having a right to know what you discuss, and the case of Greenough v Gaskell in 1883, Lord Brougham LC described it in this way:

“…if the privilege did not exist at all, everyone would be thrown upon his own legal resources; deprived of all professional assistance, a man would not venture to consult any skilful person, or would only dare to tell his counsellor half his case. If the privilege were confined to communications connected with suits begun, or intended, or expected, or apprehended, no one could safely adopt such precautions as might eventually render any proceedings successful, or all proceedings superfluous.”

As the House of Lords confirmed in 2004; ‘[privilege]…attaches to all communications made in confidence between solicitors and their clients for the purpose of giving or obtaining legal advice even at a stage when litigation is not in contemplation.’

Privilege is conferred on communications between a client and their lawyers as opposed to specifically confidential communications justified by reference to the special nature of the administration of justice. Indeed, the fact that solicitors are officers of the Court has always indicated a particular link with the administration of justice to which other professionals are not subject. A similar result is reached for barristers by reason of their Code of Conduct; and more generally nowadays by those authorised persons who are entitled to provide the reserved legal activities relating to the conduct of litigation and the rights of audience by reason of their related Codes of Conduct.

When it comes to privilege, you cannot ‘pick and choose’ what is and is not disclosed: Where the privilege in a communication has been waived, the document in its entirety, including all admissions in it, will be put before the court as per the case of Somatra Ltd v Sinclair Roche & Temperley [2000].

The High Court has an inherent jurisdiction to govern its own procedure, and part of that jurisdiction requires it to ensure that the lawyers interacting with the court conduct themselves according to proper professional standards: R (Hamid) v SSHD [2012] EWHC 3070 (Admin) (a Hamid decision). The Tribunals, Courts and Enforcement Act 2007 means the Upper Tribunal is a superior court of record, and so it can make Hamid rulings. This is what it did in November; giving a decision in the context of suspected use by lawyers of AI which resulted in false information, and particularly fake citations and authorities being placed before the Tribunal. To find out more about the use of AI in court cases, click here.

The Tribunal considered the decision of Ayinde from June 2025. For a more detailed analysis, click here. It led the Law Society to issue guidance on Generative AI. The Tribunal said that it had seen a considerable increase in the latter half of 2025 of fictitious authorities being used.  In an effort to curb that trend, the claim form has now been amended to require a legal representative to confirm by a statement of truth that any authority cited (a) exists; (b) may be located using the citation provided; and (c) supports the proposition of law for which it is cited. A legal representative who signs such a statement in a case in which false authorities should now expect to be referred to their regulatory body, so this is serious.

Additionally, the Tribunal has made clear that uploading confidential documents into an open-source AI tool, such as ChatGPT, is considered to have placed the information in the public domain, and thus to breach client confidentiality and it waives legal privilege. Everyone should therefore think carefully about the kind of tool you use and the consequences for what you put into it. If you use a tool without thought, you can lose your legal privilege, meaning information you thought was confidential will become public.

If you have questions or concerns about AI and legal privilege, please contact James Tumbridge.

To find out more about the use of AI in arbitration proceedings, click here.