AI Regulation Tracker  /  Court ruling

UK High Court sets contempt threshold for lawyers who cite fake AI-generated cases

A senior Divisional Court of the High Court in England and Wales ruled that putting non-existent, AI-generated authorities before a court can amount to contempt. The standard binds both solicitors and barristers.

By Anthony Guerriero  ·  Reviewed by The Leveraged Years Editorial Desk  ·  Published July 7, 2026  ·  Last updated July 7, 2026

A senior Divisional Court of the High Court has drawn a clear line under a run of embarrassing incidents in which lawyers filed case law that did not exist. In the consolidated matters Ayinde v Haringey LBC and Al-Haroun, the Divisional Court, presided over by the President of the King's Bench Division, Dame Victoria Sharp, set out the standard for when citing fabricated, AI-generated authorities stops being a careless mistake and becomes contempt of court.

The ruling matters because it does not simply scold the profession. It fixes a threshold. According to the court, the conduct crosses into contempt where a lawyer deliberately puts before the court material they know to be fabricated, or where, when challenged about suspect citations, they give an untruthful denial that AI was used. Both routes turn a research failure into an attack on the administration of justice.

Two cases, one standard

The two matters reached the court by different paths. In Ayinde, a judicial review over a homeless applicant's interim accommodation, the claimant's barrister put forward citations that did not exist and a summary of legislation that was wrong. In Al-Haroun, a solicitor filed a witness statement built on numerous authorities, of which, as reported in the Law Society Gazette summary, 18 were fabricated and did not exist, and others were misquoted or did not support the argument. The solicitor had relied on research supplied by his client without checking it independently.

The court's response treated both branches of the profession the same. Whether the fake authority came from a barrister drafting grounds or a solicitor relying on a client's research, the duty to verify was personal and non-delegable. Reliance on an AI tool, or on someone else's work product, did not move that duty elsewhere.

What the court did and did not decide

The court found that the threshold for contempt had been met on the facts, yet declined to launch contempt proceedings against the individual lawyers in these two cases. That restraint is the point most likely to be misread. The judges were explicit that the decision not to proceed is not a precedent and offers no shelter to the next lawyer who files invented case law. Instead of contempt, the court directed regulatory referral, sending the conduct to the Solicitors Regulation Authority for investigation.

So the ruling does not decriminalise AI hallucinations in filings. It does the opposite. It confirms that the machinery is already in place: wasted-costs orders against the lawyers responsible, referral to the SRA or the Bar Standards Board, and contempt proceedings held in reserve for the deliberate or dishonest case. The court also pressed those with leadership responsibilities in firms and chambers to put practical measures in place to stop it happening.

The verification duty in practice

For a working litigator, the operational takeaway is narrow and concrete. A generative AI tool can produce text that looks like a citation, complete with parties, court, and a plausible neutral citation, for a case that was never decided. The court's standard means the burden sits on the named lawyer to confirm, against a reliable law report or official case-law service, that each authority exists and says what the submission claims. If a citation cannot be verified, it should not be filed.

The dishonesty branch deserves separate attention. Several of these episodes escalated not because AI was used, but because the lawyer, once challenged by the court, denied using it or gave an explanation the judges found unpersuasive. That denial is what can convert a verification failure into contempt. The safer course when questioned is candour about how the research was produced.

The cross-border angle

For a US reader, this is the British counterpart to the sanctions wave that began with Mata v. Avianca, where a New York court sanctioned lawyers who filed ChatGPT-invented cases. The mechanism differs. US courts have leaned on Rule 11 and the court's inherent powers over attorney conduct, while the English court frames the exposure around contempt of court and referral to the solicitors' and barristers' regulators. The through-line is identical. A lawyer who signs or presents a filing owns every authority in it, and no AI tool absorbs that responsibility. US firms with litigation exposure in London should treat this judgment as the leading English authority and read it alongside their existing domestic guidance.

Browse the full AI Regulation News tracker

Informational analysis for working professionals, not legal advice. Confirm how any rule applies to your situation with qualified counsel.