Public AI removes Legal Professional Privilege: Munir v Secretary of State for the Home Department [2026] UKUT 81 (IAC)

The primary judicial ruling directly addressing AI and legal privilege in the UK came from the Upper Tribunal (Immigration and Asylum Chamber).

While the proceedings originally arose from lawyers submitting court filings containing AI-"hallucinated" fake case citations, the Tribunal issued significant remarks on the risks AI poses to confidentiality and privilege:

1. Open-Source vs. Closed AI Systems

• Public/Open AI Tools (e.g., public ChatGPT): The Tribunal stated that uploading confidential client documents, notes, or advice into standard consumer/public AI models is akin to placing that information on the internet in the public domain. Doing so breaches client confidentiality and can result in a waiver of legal professional privilege.
• Enterprise / Closed AI Tools: Private, enterprise-grade legal AI systems—where user inputs are not used to train the underlying models and robust contractual confidentiality protections exist pose significantly lower privilege risks.

2. Why Privilege is Waived

Legal Professional Privilege (covering both Legal Advice Privilege and Litigation Privilege) is strictly dependent on maintaining confidentiality.

• Privilege belongs to the client, not the lawyer.
• If a lawyer or client feeds privileged material into a third-party AI system where the provider gains rights over the data or uses it for model training, confidentiality is destroyed meaning privilege is lost permanently.

3. Professional Responsibility & Supervision

• Ultimate Accountability: Lawyers cannot delegate their duty of care or accuracy to AI software.
• Supervisory Duty: Supervisors are held accountable if junior fee-earners use unapproved AI tools that result in leaked privileged data or inaccurate filings. The Tribunal noted that failing to maintain these standards warrants referral to the Solicitors Regulation Authority (SRA).

Core Principles for Maintaining Privilege When Using AI

To avoid waiving privilege, UK legal practice distinguishes between compliant and non-compliant AI usage:

Regulatory & Judicial Frameworks

Alongside court rulings, regulatory and institutional bodies in the UK have set explicit guidelines:

• Solicitors Regulation Authority (SRA): The SRA confirms that while no new "AI-specific" statutes are needed, existing duties under the SRA Code of Conduct—specifically Duty of Confidentiality, Competence, and acting in the Client’s Best Interests—apply fully to AI usage.
• Judiciary of England and Wales Guidance: Official judicial guidance permits judges and legal representatives to use AI for tasks like summarizing or research, provided that client confidentiality is strictly guarded and work products are independently checked.
• Law Society Guidance: Recommends that firms adopt clear internal AI usage policies, restricting fee-earners from pasting sensitive client information or legal advice into non-vetted tools.