The rapid adoption of generative artificial intelligence tools across legal and business functions has prompted increasing judicial scrutiny. In a recent decision, the Upper Tribunal has issued a clear warning: uploading confidential or privileged material to public AI platforms can breach confidentiality and result in the loss of legal professional privilege.
This is the first time an English court or tribunal has addressed this issue directly.
The decision
In Munir v Secretary of State for the Home Department [2026] UKUT 81 (IAC), the Upper Tribunal (Immigration and Asylum Chamber) considered the professional conduct of legal representatives following concerns about the use of generative AI in case preparation. While the immediate focus of the decision was on the submission of inaccurate, AI-generated legal authorities, the Tribunal took the opportunity to make broader observations about confidentiality, privilege and the use of AI tools.
Of particular significance were the Tribunal’s comments on the treatment of confidential and privileged material uploaded to AI systems.
The Tribunal stated that uploading confidential documents into an open-source AI tool, such as ChatGPT, is to place that information into the public domain. The consequence of doing so is a breach of client confidentiality and a waiver of legal professional privilege.
Although these observations were not determinative of the appeal itself, they were expressed in clear and unqualified terms.
Why this matters
Legal professional privilege is a fundamental right under English law, but it is dependent on confidentiality. Once confidentiality is lost, privilege is lost with it, and waiver is permanent.
The Tribunal’s analysis reflects orthodox privilege principles:
- Privilege only protects confidential communications
- Disclosure to the public domain destroys confidentiality, and
- Once privilege is waived, it cannot be recovered
The Tribunal treated the act of uploading confidential material to an open-source AI tool as functionally equivalent to publishing it on the internet. On that basis, the waiver occurs at the point of upload, regardless of whether the user intended to waive privilege or believed the material would remain private.
Open-source versus closed systems
Importantly, the Tribunal drew a distinction between different categories of AI tools.
Its criticism was directed at open-source or public AI platforms, where user inputs may be stored, reused, or made accessible beyond the user’s organisation. In that context, the Tribunal considered there to be no realistic expectation of confidentiality.
By contrast, the Tribunal observed that closed, enterprise AI systems which do not place information into the public domain may be used for certain tasks, such as summarisation, without the same risks. Microsoft Copilot was cited as an example of such a tool.
While the Tribunal did not define a comprehensive test for what constitutes a “safe” system, the implication is clear: confidentiality depends on where information goes, who can access it, and whether it is retained or reused.
Status of the decision
Upper Tribunal decisions are not binding on the High Court. However, they are regarded as highly persuasive, particularly where they concern litigation practice and professional conduct.
This decision is the first judicial statement in England and Wales to address explicitly the relationship between AI use and legal professional privilege. It provides a strong indication of how courts are likely to approach the issue if it arises directly in future civil or commercial litigation.
Relationship with US case law
The Tribunal’s observations have been made against a background of recent US decisions considering similar issues, most notably United States v Heppner (S.D.N.Y., February 2026). In that case, a US federal court held that documents generated through a public AI platform were not protected by attorney-client privilege or work product protection.
While the legal tests differ, there is an important conceptual distinction between the two jurisdictions. In Heppner, the court concluded that privilege never arose because the communications were not confidential and were not made with a lawyer. In Munir, the Tribunal’s focus was on waiver: privilege existed, but was lost through disclosure to a public platform.
English courts are therefore likely to analyse AI use primarily through the lens of confidentiality and waiver, rather than treating AI tools as a substitute for legal advisers.
Practical implications for organisations
The implications of the decision extend beyond law firms. Any organisation handling confidential or legally privileged material should take note.
In practical terms:
- Confidential or privileged material should not be uploaded to public or open-source AI tools
- Internal AI policies should distinguish clearly between approved, closed enterprise systems and public platforms
- Staff should be trained to understand that convenience or efficiency does not override confidentiality obligations, and
- Particular care should be taken in contentious, regulatory or investigatory contexts, where privilege is most likely to be scrutinised
Where AI tools are used, organisations should understand how inputs are processed, stored and retained, and whether they may be accessed by third parties or used for model training.
Looking ahead
AI tools will continue to play an increasing role in legal and commercial workflows. The Munir decision does not suggest that AI should not be used, but it does underline that long-established principles of confidentiality and privilege apply with full force in the AI context.
Until there is further appellate authority, organisations should assume that placing confidential or privileged material into public AI systems carries a real and irreversible risk of privilege waiver, and should structure their AI usage accordingly.
More about the boundaries of privilege and the associated issues can be found in our analysis piece here.
