AI use and the hidden risk to legal privilege

Chloe McQuillan explores how the everyday use of generative AI tools can create an unexpected and serious risk of waiving legal professional privilege.

The use of generative AI tools is now firmly embedded in most of our day-to-day lives, including in our workplaces. From asking AI to summarise a document, to drafting and analysis, these tools offer significant efficiencies. However, their increasing use raises a fundamental issue from a legal perspective: can the use of AI tools inadvertently waive legal professional privilege (“LPP”)?

Recent judicial guidance and commentary suggest that, in certain circumstances, it can.

A reminder: privilege depends on confidentiality

Under the law of England and Wales, LPP allows parties to keep certain communications private, meaning that they do not need to be disclosed to the court or third parties in the event that a dispute arises. LPP comprises:

1. Legal advice privilege, which applies to confidential communications between a lawyer and their client for the dominant purpose of giving or receiving legal advice (for example, an email exchange between a client and their lawyer in which legal advice is requested and provided); and
2. Litigation privilege, which applies to confidential communications made for the dominant purpose of existing or reasonably contemplated litigation (for example, an expert report commissioned by solicitors after a dispute has arisen).

A central feature of both of these types of LPP is confidentiality. If confidentiality is lost (for example, if a confidential document is shared with a third party) privilege may be waived. This means that you lose protection over the previously privileged information, which could seriously impact the prospects of success of your case.

Why AI tools present a particular risk

The use of AI tools, particularly publicly available platforms, introduces significant risks to the confidential nature of privileged documents.

This is because AI tools can store, process or reuse the information that users upload to the platform. If a user uploads privileged material onto these platforms (for example, by asking Chat GPT to summarise a piece of legal advice they have received from their lawyer), this could amount to disclosure of that privileged material to a third party, which may result in privilege being waived.

This risk arises even where the user has no intention of sharing the information more widely. The very fact that the information has been uploaded to a publicly available AI platform means that the user has consented to the platform storing, processing and potentially reusing that information.

Emerging judicial approach

Although the law in this area is still developing, recent judicial decisions and commentary provide an indication of how courts are likely to approach this issue.

In England and Wales, the Upper Tribunal (Immigration and Asylum Chamber) has indicated that “[u]ploading confidential documents into an open-source AI tool, such as ChatGPT, is to place this information on the internet in the public domain, and thus to breach client confidentiality and waive legal privilege”.[1]

Similar reasoning has been applied in overseas decisions. For example, in a recent case from the United States District Court (Southern District of New York), where the judge held that legal documents produced by AI were not protected by privilege, and commented that “AI’s novelty does not mean that its use is not subject to longstanding legal principles, such as those governing the attorney-client privilege”.[2]

Distinguishing between different types of AI tools

It is important to note that not all AI use gives rise to the same level of risk.

A distinction is often drawn between:

1. Publicly available AI platforms, where inputs may be retained, analysed or made accessible beyond the user environment; and
2. Closed AI systems, which may provide contractual and technical safeguards around confidentiality.

While closed AI systems reduce the risk of inadvertent waiver of privilege, the position remains fact-specific and untested by the courts. It is therefore important to always be mindful of what information you are uploading to AI platforms, even when using closed systems.

Implications for public bodies and other organisations

The increasing use of AI tools represents a significant opportunity for enhanced  efficiency. However, it also creates a new and evolving risk in relation to legal professional privilege.

For public authorities and regulated organisations, the implications are significant. These risks include:

• Loss of privilege protection: uploading legally privileged material to AI tools may render the material disclosable in litigation, investigations, or audits;
• Enhanced disclosure risk: AI prompts, inputs, and outputs may themselves be considered to be disclosable in certain contexts; and
• Governance and compliance exposure: failure to manage AI-related risks may give rise to regulatory scrutiny, particularly where sensitive or legally privileged information is involved.

In light of these risks, organisations should ensure that they are approaching the use of AI tools (particularly in legal contexts) with appropriate caution, in order to ensure that privilege is not inadvertently lost. This should include implementing clear internal policies, training, and restrictions on uploading confidential material to AI platforms, as well as carrying out an assessment of selected AI providers’ terms and data handling practices.

Chloe McQuillan is an Associate at Sharpe Pritchard LLP.

---

For further insight and resources on local government legal issues from Sharpe Pritchard, please visit the SharpeEdge page by clicking on the banner below.

 

 

Visit Sharpe Pritchard's new Building Safety Hub, focusing on The Building Safety Act 2022 and its wide-ranging impact.

This article is for general awareness only and does not constitute legal or professional advice. The law may have changed since this page was first published. If you would like further advice and assistance in relation to any issue raised in this article, please contact us by telephone or email This email address is being protected from spambots. You need JavaScript enabled to view it..

[1] UK v Secretary of State for the Home Department (AI hallucinations; supervision; Hamid) [2026] UKUT 81 (IAC) [60].

[2] United States v Heppner, No 25 Cr 503 (JSR) (SDNY, 17 February 2026) (Rakoff J).