UK courts face structural challenge from AI-hallucinated evidence

TL;DR:

• Disputes lawyer Nicholas Blomfield argues AI-hallucinated evidence is shifting from one-off scandal to routine procedural risk, with England and Wales’ civil justice system unprepared for the scale.
• The flagship case — R (Ayinde) v London Borough of Haringey — saw fabricated case law submitted to the High Court, a wasted-costs sanction imposed and both barrister and solicitor referred to their regulators.
• Blomfield calls for mandatory certification by legal representatives that AI outputs have been verified, plus adverse inferences where AI use is concealed.

The opinion lands as the British legal profession works through how to absorb tools such as Harvey AI and Legora into mainstream practice. Blomfield’s central claim is that the Civil Procedure Rules were built around human accountability — a witness signs a statement of truth, a solicitor certifies documents — and that probabilistic autonomous systems sit outside that framework.

Context and Background

The Ayinde judgment, in which Dame Victoria Sharp P warned that unverified reliance on AI constitutes negligent conduct, has become the reference case for UK legal-AI governance. The IOPC’s separate report this week — covered by Resultsense — that a quarter of judicial review requests are now likely AI-drafted, some citing invented English statutes, suggests the problem is no longer confined to one-off court appearances.

Blomfield highlights three procedural pressure points: authenticity (courts know how to distinguish genuine documents from forgeries, but not the new “hybrid synthetic” category); attribution (the platform, the solicitor or the client — under existing rules, responsibility is hard to pin); and weighting (how should a court treat material partly generated by a system capable of “authoritative but false” output?). A boilerplate disclosure that AI was used “for drafting assistance only” is, he argues, meaningless without substantive explanation of how the system shaped the content.

For UK in-house legal teams and law firms, this is the second time in a fortnight that AI-hallucinated evidence has been treated as a structural risk rather than user error — the High Court and the police watchdog now broadly aligned on the diagnosis.

Looking Forward

Blomfield’s call for mandatory verification certificates is unlikely to be adopted by individual judges case by case; it would more plausibly land via a Civil Procedure Rule Committee update or Bar Standards Board guidance. For now, UK firms should expect AI-use disclosure obligations to harden through 2026, with growing professional-conduct consequences for failure to verify generative output before submission. Resultsense will track the procedural response as the Ministry of Justice and the senior judiciary catch up with the evidential reality.