AI-Generated Legal Research: Royal Courts of Justice Stress Duty of Lawyers to Verify Authorities

The Royal Courts of Justice, in the conjoined matters of R (Ayinde) Vs London Borough of Haringey and Al-Haroun Vs Qatar National Bank QPSC & Anr., has delivered an important judgment on the professional obligations of lawyers who use generative artificial intelligence tools in litigation.

The Divisional Court, sitting in the High Court of Justice, King’s Bench Division, examined two Hamid jurisdiction referrals arising from the use – or suspected use – of generative AI to produce court documents containing fabricated case law, distorted statutory extracts, and unsound legal propositions. The ruling is a significant marker for how courts expect legal professionals to handle AI tools such as ChatGPT when preparing pleadings, skeleton arguments, and evidence.

The court accepted that artificial intelligence can contribute meaningfully to litigation – particularly in areas like document review and disclosure. However, it underscored that public, freely available generative AI models trained on large language datasets are not capable of reliable legal research.

Key warning: AI-generated text may look authoritative but can:

  • Invent case names and neutral citations
  • Attribute fictional quotations to genuine cases
  • Misstate statutory provisions and legal principles with apparent confidence

Accordingly, lawyers remain under a non-delegable professional duty to ensure that any legal authority or proposition placed before a court is checked against authoritative sources such as:

  • Official legislation databases maintained by the Government
  • Judgments preserved by the National Archives
  • The official Law Reports published by the Incorporated Council of Law Reporting
  • Reputed commercial legal research platforms

This obligation applies regardless of whether the material originates from:

  • An AI tool
  • An intern, paralegal, trainee solicitor, pupil barrister
  • A lay client’s own “research”

The court stressed that AI does not dilute or replace the professional responsibilities of advocates and those conducting litigation.

Professional Guidance Already in Place – And Its Limits

The Divisional Court surveyed the current landscape of professional guidance on AI use in legal practice, noting a clear, consistent message: AI outputs must never be relied upon without verification.

Existing Guidance to the Bar

The Bar Council’s January 2024 document, “Considerations when using ChatGPT and generative artificial intelligence software based on large language models”, was expressly cited. It highlights that:

  • Large language models can produce convincing but false content.
  • Barristers must not accept such outputs at face value.
  • Misleading the court, even inadvertently via unverified AI content, can amount to conduct that is incompetent and grossly negligent, potentially breaching Core Duty 5 (not to bring the profession into disrepute).
  • Barristers risk disciplinary action, professional negligence claims, defamation, and data protection issues if they carelessly use AI outputs.

The guidance explicitly requires barristers to verify AI-generated outputs and to maintain robust checking procedures.

Guidance for Solicitors

The Solicitors Regulation Authority’s “Risk Outlook report: the use of artificial intelligence in the legal market, 20 November 2023” similarly warns that:

  • AI language models such as ChatGPT operate by predicting text, not understanding reality.
  • This phenomenon leads to “hallucinations” – highly plausible but incorrect or fabricated content.

Solicitors are therefore reminded that they remain accountable for what they present to the court and to their clients, irrespective of whether AI assisted in its creation.

Bar Standards Board Commentary

A Bar Standards Board blog post dated 8 October 2023, “ChatGPT in the Courts: Safely and Effectively Navigating AI in Legal Practice”, referred to the US case Mata v Avianca Inc. In that matter, lawyers had submitted pleadings drafted with ChatGPT that contained entirely fabricated authorities and spurious quotations. The blog emphasises:

AI can assist, but it cannot replace human responsibility and oversight.
The barrister is answerable for research, arguments, and representations under the core duties to the court and client.

Judicial Guidance – Also Relevant to Practitioners

The judiciary has separately issued guidance (first in December 2023 and revised in April 2025) on AI use by judges, which is equally instructive for lawyers. That guidance stresses:

  • Confidential or non-public information must not be entered into public AI tools.
  • Any AI-generated information must be verified before use.
  • AI tools may:
    • Invent cases, citations, or quotes
    • Refer to legislation or texts that do not exist
    • Misstate the law or its application
    • Introduce factual inaccuracies

The judicial guidance sums up the essential principle:

“All legal representatives are responsible for the material they put before the court/tribunal and have a professional obligation to ensure it is accurate and appropriate.”

The guidance also calls out that public AI chatbots are poor instruments for new legal research and do not produce reliable legal analysis.

Regulatory Duties of Barristers and Solicitors

The court carefully tied AI misuse back to established regulatory frameworks.

Barristers’ Obligations

Under the Bar Standards Board Handbook, and its Code of Conduct:

  • Barristers owe a duty to the court in the administration of justice (CD 1).
  • They must act with honesty and integrity (CD 3).
  • They must not reduce public trust in the profession (CD 5).
  • They must deliver a competent standard of work (CD 7).

Outcomes sought by these rules include:

  • The court can rely on information supplied by advocates and those conducting litigation (Outcome 1).
  • The proper administration of justice (Outcome 2).
  • Clarity among advocates as to their duties to the court (Outcome 4).

Specific rules require barristers to:

  • Not knowingly or recklessly mislead the court (Rules C3.1, C9.1).
  • Avoid drafting documents that make contentions they do not consider properly arguable (Rule C9.2.b).
  • Provide competent work (Rule C18).