Series
Blogs
A wAIver of privilege? The use of AI and the implications on legal professional privilege
A wAIver of privilege? The use of AI and the implications on legal professional privilege
5 May 2026
Series
Blogs
5 May 2026
Authors: Sinead Casey, Leanne Raven
AI tools are increasingly deployed across a range of legal tasks, including document review and diligence, contract drafting and review and legal research. With the ease of using generative AI tools there is an increased risk of inadvertent waiver of legal professional privilege, leaving employers exposed. We consider recent developments in this area and best practice for employers to mitigate these risks.
Privilege entitles a party to withhold evidence from production to a third party or the court. There are a number of types of privilege, including legal professional privilege which is made up of:
It is clear that the concept of “confidentiality” is key to both definitions, meaning that a communication which is not confidential cannot be privileged. It is also important to note that once privilege is waived, it cannot be restored.
We do not yet have authority from the Employment Tribunal about the interaction between AI and legal professional privilege. However, the recent combined judgment of the UK v Secretary of State for the Home Department [2026] UKUT 00081 in the Upper Tribunal Immigration and Asylum Chamber has shed some light on judicial thinking in this area. Whilst the decision dealt with two cases of legal representatives relying on fictitious authorities as a result of using AI, the Judge also provided a helpful view on AI use and the impact on confidentiality and privilege, stating:
“Uploading 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, and any such conduct might itself warrant referral to the SRA and should, in any event, be referred to the Information Commissioner’s Office.” The Judge also added that “Closed source AI tools which do not place information in the public domain, such as Microsoft Copilot, are available for tasks such as summarising without these risks.”
The Judge’s use of the terms “open-source AI tools” and “closed source AI tools” are at odds with how the terms are used in the AI-sphere (whereby “open source” typically refers to freely available source code which is released openly to the public vs closed source models which are proprietary systems that keep their code confidential). However, the judgment does provide some insight into judicial opinion on the topic of uploading data to different types of tools. It indicates that where privileged information is uploaded to a tool such as publicly available ChatGPT or Claude, the tool may use the input and/or output data to further train its model, improve it or provide it to third parties. In these situations, confidentiality is not preserved and privilege may be lost.
If, however, such privileged information is uploaded to an AI tool, such as an enterprise AI tool used by an employer, which is private, prohibits training on any data uploaded, prohibits sharing data with any third parties and maintains data in a safe and secure environment - this may be less likely to cause privilege issues (depending, of course, on the specifics of the tool and terms of the platform). However, this does remain untested. Whilst this can provide some comfort to employers deploying enterprise AI, they still need to be alive to the risk of employees using “shadow AI”. Shadow AI is the term used to describe the use of an unapproved AI tool by employees. In the workplace context, this may be that an employee only has access to Copilot but they much prefer ChatGPT. Accordingly, they use ChatGPT on their personal phone inputting confidential information. Such use of shadow AI can clearly pose a material risk to confidentiality and privilege.
Crucially, even where communications are not privileged, it is worth a reminder to staff that their prompts, conversations, inputs and outputs may all be surfaced in a disclosure exercise. The manner in which people interact with generative AI is often fundamentally different (sometimes more personal, transparent or conversational) to how they interact with other forms of technology (for example searching on the internet or writing an email). When interacting with such generative AI tools in the workplace, it can help focus employees’ minds to ask them to consider how they would feel if their prompt or chatbot conversation was discovered in a disclosure exercise and read out in Tribunal.
Does obtaining legal advice from an AI tool attract legal professional privilege? The degree of lawyer involvement is likely to be critical: the more AI operates autonomously, the harder it may be to characterise the output as a privileged communication from a legal adviser.
This was one of the questions the US courts have recently grappled with in the case of United States v Heppner, 25 Cr. 503 (JSR) in what appears to be the first decision of its kind nationwide. The defendant, facing fraud charges, had used the AI platform Claude to prepare documents outlining his defence strategy and had subsequently shared them with his lawyers. He claimed privilege over these documents. The Court rejected this because:
There were also arguments as to why the “work product” doctrine failed. The “work product” doctrine essentially protects materials prepared by or at the direction of counsel in anticipation of litigation. Quite simply, in this case the documents were not prepared by counsel or at counsel’s direction but by the defendant on his own volition.
Whilst the US concepts of privilege do not map precisely onto the English law concepts, the decision serves as reminder that clients who use AI platforms to assist their own legal thinking, particularly without their lawyers' direction, may inadvertently place sensitive material beyond the reach of any privilege protection, rendering it potentially disclosable.
Closer to home, a recent High Court decision handed down in April 2026 was significant in potentially expanding the application of legal advice privilege to apply to “intra-client” communications (i.e. internal communications between members of the client group).
Specifically, the Judge held that a party is “entitled to assert legal advice privilege in respect of intra-client documents provided that those documents were created with the dominant purpose of seeking legal advice”. This potentially expands the scope of privilege by providing that legal advice privilege can apply to communications and documents created by or sent between members of the relevant client group for the dominant purpose of seeking legal advice, even where there is no lawyer in the loop, rather than solely covering communications between lawyers and their clients made for the dominant purpose of seeking or giving legal advice.
As a High Court decision this is not binding on other courts. Whilst this particular judgment is not addressing AI, looking at the concept of privilege through an AI lens, our advice is that clients using generative AI to create documents with the dominant purpose of seeking legal advice should exercise caution and continue to work on the basis that the inputs and outputs would not attract legal advice privilege.
Against this backdrop employers should be considering taking the following steps.
We are advising clients on the risks that arise with use of AI in the workplace. If you have any questions on this article or related topics, please do get in touch with our experts.