SRA v Kumar (Solicitors Disciplinary Tribunal)

What happened

The Solicitors Regulation Authority brought proceedings against Abhishek Kumar, a Registered Foreign Lawyer, on an allegation arising from a 2024 criminal conviction that has nothing to do with artificial intelligence. Kumar represented himself. On March 12, 2026, he filed his Answer to the SRA's statement.

The Answer cited authorities that did not exist, gave wrong citations, misquoted principles, attributed passages to judgments that do not contain them, and cited cases for propositions they do not support. The SRA wrote to him, drew his attention to the Divisional Court's judgment in Ayinde v Haringey, and asked whether he had used AI. In an email of April 9, 2026, he admitted using AI to draft the Answer and to find case law. That email itself contained further wrong citations and quotations; among other things it attributed to SRA v James passages the SRA could not find in the judgment. On April 13, 2026 he accepted that the email had also been drafted with AI. The SRA then brought two further allegations based on the AI-drafted material.

What the court did

After a hearing on July 29, 2026, the Tribunal found both AI allegations proved, as breaches of Principles 1, 2 and 5 of the SRA Principles and paragraphs 1.4 and 2.4 of the Code of Conduct, which prohibit misleading the court and require that submissions be properly arguable. Dishonesty was not alleged. The Tribunal held that typographical errors in citations were not serious enough to amount to misconduct, but the citing of nonexistent cases and incorrect legal propositions was.

The judgment records that "This was the first time that a lawyer's use of artificial intelligence or large language models in legal proceedings had been litigated before the Tribunal." On sanction, the Tribunal assessed culpability as very high because the conduct "was repeated after errors had been pointed out and explained," and it gave "very substantial weight to the repeated nature of the Respondent's misconduct." It rejected the submission that acting for himself rather than a client was mitigating, because he sought to be the direct beneficiary of the conduct.

The Tribunal ordered him struck off the Register of Foreign Lawyers, with no order as to costs. It stated that "the sanction it reached was the same sanction it would have imposed if either allegation had appeared before the Tribunal without the other."

Where it stands

The judgment is dated August 25, 2026. No appeal to the High Court was located as of October 8, 2026. Because the order also rests on the separate conviction allegation, the strike-off is not an outcome of the AI conduct alone, although the Tribunal said it would have reached the same sanction on the AI allegations by themselves.

SRA v Kumar is the regulatory follow-through that the Divisional Court invited in Ayinde, where it asked the legal regulators to consider what steps beyond guidance were needed. The Tribunal's distinction between harmless citation slips and nonexistent authority, and its emphasis on repetition after warning, are the two points most likely to be cited in later proceedings.

The obligation it engaged

The conduct engages Clause 3 of the Standard, independent authority. Every authority and quotation in the Answer had to be confirmed against the judgment it came from; the SRA did that check and the respondent did not. Clause 4, checkability before adoption, is engaged by what happened next: after the errors were identified, the next document was produced the same way and contained the same kind of errors.

The Standard is written for experts, but the lesson carries over directly. An expert who learns that a system has supplied a false citation has learned something about the method, not just about one citation, and the record of what was checked needs to change with it.

Sources

• Solicitors Disciplinary Tribunal (copy hosted for the AI Hallucination Cases database) · August 25, 2026
  Judgment, Case No. 12884-2026
• Damien Charlotin, AI Hallucination Cases database · October 8, 2026
  AI Hallucination Cases

What could not be confirmed

The judgment does not name the AI tool. The copy read was the database's hosted copy; the Tribunal's own website could not be reached from the research environment. No appeal was located as of 2026-10-08.