Generative AI tools have become common in legal practice, but courts are rejecting claims that AI hallucinations excuse fabricated case law. Recent rulings across the UK, Canada, and Australia establish that lawyers cannot rely on AI-generated fabrications as a defense for submitting false authorities to court.
The UK’s first direct confrontation with AI-generated legal fabrications came in Harber v Commissioners for HMRC [2023] UKFTT 1007 (TC). A self-represented appellant in a tax penalty appeal submitted fabricated authority citations that she acknowledged may have been generated using ChatGPT. The tribunal found neither HMRC nor court officials could locate the cited cases, which contained fictitious judge names, quotes, and legal outcomes.
The First-tier Tribunal rejected the fictitious authorities and provided guidance on how genuine decisions could be independently verified. The ruling established two key principles: AI-generated fabrications carry no weight in judicial proceedings, and while self-represented litigants might receive initial leniency, licensed practitioners should expect no such clemency.
Professional Standards Tightened
By 2025, the UK High Court set clear expectations for qualified practitioners in Ayinde v London Borough of Haringey; Al-Haroun v Qatar National Bank [2025] EWHC 1383 (Admin). Acting counsel presented fabricated authorities in court submissions, raising serious questions about generative AI use. Unlike Harber, the court dealt with admitted legal professionals bound by professional practice and ethical obligations.
The judgment established a Responsible Legal Augmentation framework requiring independent verification of authorities, prohibiting fabricated citations as professional conduct breaches, upholding the duty to the court, maintaining human accountability, and exercising professional competence. The ruling effectively dismantled the “hallucination defense” for legal counsel and firms, confirming that ignorance of LLM limitations does not excuse failures to verify submitted material.
Related Post: Direct access becomes standard route for legal services
The judicial response intensified in Munir v Secretary of State for the Home Department [2026] UKUT 81 (IAC). The Upper Tribunal addressed AI hallucinations, duty of honesty in judicial review proceedings, and the tribunal’s supervisory jurisdiction under Hamid principles. The court warned that submitting hallucinated authorities in public law claims constitutes serious abuse of court process, with potential regulatory referrals to bodies like the Bar Standards Board or Solicitors Regulation Authority.
In Zhang v. Chen, an application in the Supreme Court of British Columbia, counsel unwittingly included non-existent case citations generated by ChatGPT in a family law matter. Upon discovery, counsel promptly withdrew the fake cases and apologised, explaining she was unaware of ChatGPT’s hallucination risks. The court held that, although using AI for legal research is not inherently improper, citing fictitious authorities is a serious error that causes opposing counsel and the court to expend substantial time and resources investigating and correcting the resulting confusion.
While the court declined to award special costs in this case, it noted that such conduct could expose counsel to personal special-costs liability. Counsel was nevertheless ordered personally to bear the additional ordinary costs caused by the false citations and the delay in correcting them, and to review her other court files for similar material.
Common Law Consensus and Regulatory Scrutiny
Australian courts and tribunals have issued practice notes addressing generative AI use, emphasizing that the duty owed by legal practitioners to the court remains non-delegable. Submitting hallucinated authorities constitutes a breach of the practitioner’s personal warranty that cited materials are accurate and relevant, regardless of intent to mislead.
Professional Liability and the Duty of Human Verification
Legal probabilistic models, including both specialist platforms like Harvey and Legora and general-purpose large language models, can produce plausible yet fictitious case law, statutory provisions, and judicial reasoning. This tendency to “hallucinate” creates a persistent vulnerability when such outputs are incorporated into legal research, drafting, or advice.