UK v Secretary of State for the Home Department [2026] UKUT 81 (Hamid)

In brief

The Upper Tribunal set out the professional duty not to cite false cases and the privilege risks of uploading material to public AI tools. It made no further referral, recording that "We make no referral to the IAA or SRA as Mr Mohammed has already done this and there is no need for us to do so." The judgment's force lies in its statement of the professional obligation, holding that those who cite false cases fail to comply with it and waste the Tribunal's time, and that a supervisor who fails to prevent a junior's fabricated citations "is likely to be more culpable" than one who fails to check only their own work. First English/Welsh tribunal or court decision directly addressing the implications for legal professional privilege when confidential or privileged material is uploaded to AI tools. The Upper Tribunal (Immigration and Asylum Chamber) drew a critical distinction between public/consumer-tier AI tools (the Tribunal used the words 'open source AI tools such as ChatGPT') — uploads to which 'place this information on the internet in the public domain' and waive privilege — and closed enterprise/network-isolated AI tools (citing Microsoft Copilot) which 'do not place information in the public domain' and are 'available for tasks such as summarising without these risks'. Reinforced the supervisory obligation per R (Ayinde) [2025] EWHC 1383.

Key provisions

• AI / privilege distinction — Public/consumer-tier AI uploads waive privilege; closed enterprise AI does not : Uploading confidential or privileged material to public-tier AI tools (e.g. ChatGPT) places that information in the public domain and waives privilege. Closed enterprise/network-isolated AI tools (e.g. Microsoft Copilot configured at enterprise tier) do not have this effect and are available for tasks such as summarising.

When relevant

Cases involving AI tooling in legal/governance work; engagement with instructing solicitors on AI use; diligence on consultant/expert AI practices.