By Anthony Dique and Snehal Patel
A recent English Upper Tribunal decision in UK v Secretary of State for the Home Department [2026] UKUT 81 (IAC) has once again thrown the spotlight on the use of Artificial Intelligence (AI) within the legal profession.
While it should not be surprising to readers that AI has once again hallucinated citations and authorities, which have once again been submitted to a Court, this decision appears to be the first instance where it was noted that the use of an “open-source” AI tool, such as Chat-GPT, would breach client confidentiality and waive legal privilege.
Case overview
The Judgment arose from the Upper Tribunal considering a solicitor’s inappropriate uses of AI software which resulted in fake citations and authorities being placed before the Upper Tribunal.
The solicitor had drafted and filed grounds of appeal to the Upper Tribunal in a migration matter, which cited a case described as Horleston v SSHD [2007] EWCA Civ 654. The Upper Tribunal could not locate the case and asked the firm on the record to please explain.
Initially, the solicitor described the false citation as an inadvertent human error and denied any AI use, but later admitted that he could not explain the error from his browser history and that, given how AI works, it was likely an AI creation arising from inadvertent use of Google’s “AI mode”.
The solicitor further admitted that, while he had not used AI to draft his legal grounds, he had used ChatGPT to:
- improve client emails explaining Home Office decisions.
- summarise Home Office decision letters by uploading them to the platform.
The solicitor self-referred himself to the legal regulatory body in the UK.
Decision
The Upper Tribunal accepted that the use of legal AI programs by properly trained professionals is a step forward in legal practice, but warned of the risks of using a “non-specialist” AI tool to undertake such legal research or drafting.
The Upper Tribunal pointedly noted that it cannot afford to have its limited resources absorbed by representatives who place false information before the Tribunal.
The Upper Tribunal also remarked with concern that to put client information and confidential information into an “open-source” AI tool, such as Chat-GPT, is to place client and confidential information on the internet in the public domain, and would amount to breaching client confidentiality and waive legal privilege.
The Upper Tribunal noted that “closed source” AI tools which do not place information in the public domain are available for tasks without these risks.
Conclusion
This case serves as not only another reminder for legal practitioners navigating the risks associated with AI in legal practice, but also a fresh warning about the risks of uploading client information into AI tools.
