In a landmark decision, the Solicitors Disciplinary Tribunal (SDT) has ordered that a registered foreign lawyer be struck off after he submitted fabricated, AI-generated legal authorities to the regulator and tribunal. The judgment marks the first instance in the United Kingdom where professional striking off has been imposed for generative AI hallucinations in formal legal submissions.
The misconduct: Double hallucination
During regulatory proceedings brought by the Solicitors Regulation Authority (SRA), the practitioner relied on unverified generative AI tools to draft his defense representations. The submissions cited entirely fictitious cases, false neutral citations, and fabricated judicial holdings presented as genuine English precedent.
When challenged by the regulator regarding the authenticity of the authorities, the practitioner compounded the misconduct by using generative AI a second time to draft an explanatory statement—which itself contained further fabricated details and misleading statements.
Tribunal findings and standard of care
The SDT found that the lawyer’s culpability was “very high,” emphasizing that legal practitioners owe an uncompromised duty of candour and accuracy to courts and regulatory bodies. The tribunal ruled that:
- Practitioners remain personally liable for every citation, authority, and factual representation submitted under their name, regardless of whether AI tools assisted in drafting.
- Blind reliance on AI outputs without verifying primary legal sources undermines the administration of justice and amounts to misleading the tribunal.
- Sanctions for knowingly or recklessly submitting unverified hallucinations will be severe, with striking off warranted to maintain public trust in the profession.
Practical implications for legal practitioners
Following this precedent and the SRA’s recent Warning Notice on the Misuse of Artificial Intelligence, law firms and in-house counsel operating in England & Wales must enforce strict compliance controls:
- Mandatory source verification: Never cite or rely upon case law, statutory provisions, or secondary sources generated by AI without verifying them against authoritative law reports (such as ICLR, BAILII, or Westlaw).
- Confidentiality safeguards: Entering client data or privileged instructions into consumer-grade, unmanaged AI models creates unacceptable regulatory risks under the SRA Code of Conduct.
- Firm-wide AI governance: Implement clear internal policies, audit trails, and training to ensure junior associates and fee-earners understand the operational boundaries of legal generative technology.