AI and the challenges of maintaining legal privilege

AI has had a profound impact on many aspects of commercial life, including the way that parties to civil disputes draft, review and manage documents.

AI generated documents are increasingly common and offer both business and individual litigants clear advantages in terms of efficiency and accuracy.

However, these changes to the litigation landscape are not without their challenges.

In this article, we look at the potential risks to the concept of legal privilege posed by the ever increasing use of AI.

What do we mean by legal privilege?

The concept of legal privilege is one of the cornerstone of English law.

It protects confidential communications between a lawyer and their client from disclosure, ensuring that clients are able to obtain legal advice without fear that the contents of that communication will need to be divulged to and potentially exploited by another party.

Advantages of Legal Privilege

Legal advice privilege

This form of privilege protects confidential communications between a client and their lawyer where the dominant purpose of that communication is to obtain or give legal advice. It applies to both written and oral communications and extends to documents created as part of the process of obtaining legal advice.

Where it applies, those communications are not required to be disclosed to third parties or the court.

For the purposes of legal advice privilege, the term “client” is relatively narrowly defined. Where the party seeking or receiving the legal advice is a company, it would not be unreasonable to assume that the client is the company as a whole.

However, the English courts have taken a much more limited view. The client is only those individuals within the organisation who are tasked with obtaining legal advice from lawyers on the matters in question.

This means that businesses must take care to clearly identify which individuals are authorised to engage lawyers and to manage internal communications disseminating that advice as privilege does not apply as widely as often is assumed.

Litigation privilege

1. Litigation must be in progress or reasonably in contemplation. The mere possibility that there could be litigation is in the future is not sufficient. There has to be a real prospect of litigation at the time the communication or document is created.
2. The communication or document must be created for the dominant purpose of the litigation. This means that the main or overriding reason a document was created, or a communication exchanged, is to conduct or prepare for existing proceedings or proceedings that are reasonably in contemplation. Do not assume that just because litigation is in existence that any and all documents created during that period that have any form of connection with the issues in dispute, however tangential will automatically attract litigation privilege. For example, documents that are produced for general fact finding purposes, or for the purposes of commercial decision-making or internal management, will not necessarily be covered. It will come down to the dominant reason for bringing those documents into existence.
3. The communication must be kept confidential. This means that the sharing of the communication or document must be restricted to the client, the lawyers and any relevant third parties (such as expert witness). Sharing it more widely means that that it may cease to be confidential, which may result in privilege being lost or waived.

The use of AI in document creation and management

AI tools are commonly used to assist in the drafting of commercial contracts, preparing legal advice, reviewing evidence and managing large-scale disclosure exercises as part of the litigation process.

AI undoubtedly offers significant benefits in terms of making processes more efficient and controlling costs, while allowing the human element to focus more on the strategic direction of the case.

But as the use of AI becomes more widespread, questions are being asked about the risks posed by AI generated documents to ensure that public confidence in the administration of justice is not undermined.

One only needs to look at the ever increasing number of cases of AI hallucinations in this jurisdiction and abroad to understand the risks posed if the use of AI is allowed to continue unchecked. While there are no official statistics, the number of reported cases in the UK involving hallucinated case citations (i.e. fake or fabricated legal authorities) as of early 2026 is believed to be between 35 and 50 individual cases.

This has resulted in bodies such as The Civil Justice Council (“CJC”) consulting on the use of AI in the preparation of court documents.

One of the suggestions that has been put forward is restricting the use of AI in the drafting of witness statements. The Civil Procedure Rules state that a witness statement must be in the witness’ own words and even where a lawyer assists with drafting the statement, it must be based on and not go beyond any note or record of the witness’ own version of events.

It is not the role of the lawyer to embellish or add their own gloss to a statement.

The CJC has expressed concerns about how that rule can be upheld if AI is used to produce the statement. It has therefore proposed that going forward lawyers will be required to make a statement that AI has not been used for the purposes of generating the content of the statement.

Is AI generated legal advice privileged?

When a dispute arises, it is becoming increasingly common for individuals within businesses to use AI to obtain preliminary legal advice on a particular issue. Lawyers may not play any part at this stage.

If that advice is then shared will it be covered by legal privilege? This seems highly unlikely based on the current status of English law.

First, for legal advice privilege to apply there needs to a confidential communication between the client and their lawyer. The Supreme Court has held that in this context a lawyer is a qualified legal adviser acting in their professional capacity.  This includes solicitors, barristers and in-house counsel, but does not extend to accountants, tax advisors, or lawyers acting in a non-legal capacity.

This definition suggests that a lawyer can only be a natural person, which on current definitions would exclude AI systems on the basis that they have neither legal personality nor human consciousness.

This view is supported by the Supreme Court decision in Thaler v Comptroller-General of Patents, Designs and Trade Marks [2023] UKSC 49. This case concerned two applications for the grant of patents where the processes to be patented were generated by an AI powered machine called “DABUS”.

The Comptroller-General formed the view that the applications could not proceed on the basis that AI did not qualify as an inventor within the meaning of the Patents Act 1977 and a natural person could not claim to be the inventor simply because they owned the machine. The Supreme Court ultimately upheld that decision and found that:

… an inventor within the meaning of the 1977 Act must be a natural person, and DABUS is not a person at all, let alone a natural person: it is a machine … [and] only persons with legal personality can be granted a patent, not machines.

If a machine cannot be an inventor for the purposes of the Patent Act 1977, it is difficult to see how an AI tool could be held to be a lawyer for the purposes of attracting legal advice privilege.

That is not to say, however, that the position could not change. In February 2017, the European Parliament requested that the Commission consider options for assigning legal personhood to self-learning robots with a view to allowing them to be held liable in the event of damage to persons or property.

This may sound fanciful and ultimately it was rejected by the European Parliament, but it does demonstrate that the law is not immutable and will adapt to ever-changing technology advancements.

Making use of AI to obtain legal advice

We are seeing an ever increasing number of emails from potential clients requesting legal advice that have obviously been drafted by or with the assistance of AI. This is nothing intrinsically wrong with that, but does the involvement of AI impact on its status as a privileged document?

The use of an AI tool to draft an email to a lawyer for the purpose of obtaining legal advice is likely to still be covered be legal advice privilege. This is because the email, although generated by the AI tool, is the result of input from the individual and provided it was prepared for the purpose of obtaining legal advice there is no obvious reason why privilege would not apply.

The same is likely to apply to a lawyer who has used AI to produce an email of advice to their client.

It is important to note that there do not yet appear to have been any published judgments of the English courts where this specific issue has been definitively decided.

Making use of AI generated legal advice

Another common occurrence is where AI is used to summarise a longer document or provide action points arising out of that document. What is the position if that document happens to be a written legal opinion provided by a lawyer? If that advice is then shared with a public AI tool, is there a risk that privilege in that advice may then be lost?

The potential risk posed by uploading legal advice to the likes of ChatGPT were highlighted by the guidance issued to judicial office holders in England and Wales in April 2025. It expressly warned against entering any information that is private and confidential into a public AI tool on the basis that such action “should be seen as being published to all the world.”

This is because AI tools remember the information entered into them and that information is then used to respond to queries from other users. As a consequence, any information inputted or uploaded to a public AI tool may become public knowledge.

Judges have been warned that they “should treat all public AI tools being capable of making public anything entered into them.”

If those concerns hold true, then there is a very real risk that uploading confidential legal advice into a public AI tool, such as ChatGPT, could result in privilege in that advice being lost and it potentially becoming a disclosable document.

Our recommendation is that legal advice should never be uploaded into a public AI tool.

If the AI model used is closed or proprietary model and not accessible by the general public, then it may be possible to maintain confidentiality in that advice provided that access is restricted to those who need access to it. However, this issue has not yet been tested by the English courts and is not without risk.

An example from the United States

In United States v Heppner, the court held that materials generated by Claude were not protected by client-attorney privilege. It appears that the defendant, who has been charged in connection with an alleged fraud, had uploaded information into Claude to assist him in preparing in legal response to the case.

Federal agents identified 31 recorded exchanges with Claude. When discovery of those documents was sought, counsel argued that they were privileged on the basis that they had been produced for the purposes of assisting with the litigation.

This was rejected by the Court, which held that because confidentiality had been lost because the information had been uploaded into an open model AI tool.

Although not binding on the Courts in this country, it may give an insight into how the Courts of England and Wales may approach this issue.

Use of AI for transcribing calls

The use of AI for note taking during conference calls is now commonplace. However, if the purpose of that call is to discuss legal advice and a lawyer is not present, will the transcript of that call be privileged?

Legal advice privilege is unlikely to apply and it may not be covered by litigation privilege if litigation is not reasonably in contemplation at the time of production of the transcript.

Caution should always be exercised, because if privilege does not apply and court proceedings are subsequently commenced, the transaction may be disclosable.

Use of AI in Court proceedings

The potential risks associated with the use of AI in litigation were addressed by the Divisional Court in the matter of The King on the application of Frederick Ayinde v The London Borough of Haringey [2025] EWHC 1383 (Admin).

The Court recognised that AI is a tool that “carries with it risks as well as opportunities.” The judgment identifies that “[freely] available generative artificial intelligence tools, trained on large language models such as ChatGPT are not capable of conducting reliance legal research. Such tools can produce apparently coherent and plausible responses to prompts, but those coherent and plausible responses may turn out to be entirely incorrect. The responses may make confident assertions that are simply untrue. They may cite sources that do not exist. They may purport to quote passages from a genuine source that do not appear in that source.”

This raises serious implications for the administration of justice.

This does not mean that AI tools cannot be used for preparing documents for litigation, or conducting research, but the outputs should not necessarily be accepted without critical thinking.

The onus will be on the party filing that document or preparing their submissions or legal arguments to the court, to check that the information put before the court is accurate. The failure to do so can result in severe consequences for the parties concerned.

Conclusion

AI offers business and lawyers, many potential benefits. But that does not mean that AI could be used without thought to how that use might jeopardise core concepts of English law such as legal advice privilege.

Parties must continue to take proper steps to ensure that legal advice is properly protected by ensuring that confidentiality in that advice is maintained.

The consequences of failing to do so could be severe.

How can Morr & Co help?

If you would like more information and advice on the use of AI in legal professions or in any other business. Please do not hesitate to Contact our Dispute Resolution team on 0333 038 9100 or email info@morrlaw.com and a member of our expert team will get back to you.

Disclaimer
Although correct at the time of publication, the contents of this newsletter/blog are intended for general information purposes only and shall not be deemed to be, or constitute, legal advice. We cannot accept responsibility for any loss as a result of acts or omissions taken in respect of this article. Please contact us for the latest legal position.