Court Finds Fake Citations Blamed on AI Were Human Error
A court probe found that paralegal editing—not AI—caused fabricated legal citations in filings.
Why it matters: The finding challenges the assumption that AI is solely responsible for citation errors in legal work. Law firms must reinforce human oversight and quality control, even when AI tools are used.
- Court investigation attributed fake citations to human paralegal editing, not AI.
- Disciplinary action has targeted attorneys who submitted briefs with bogus AI-attributed citations.
- U.S. courts cite 95 cases since June 2023 of fake research in legal filings, often linked to AI use.
- Sanctions and warnings over fake citations continue on both sides of the Atlantic.
Legal tech's growing pains are under the microscope after a court determination that a much-publicized batch of bogus legal citations, initially attributed to AI hallucination, actually stemmed from sloppy paralegal editing. The finding undermines the now-familiar narrative that AI is solely behind citation errors plaguing court filings.
- The State Bar of California recently filed disciplinary charges against attorneys Omid Emile Khalifeh and Steven Thomas Romeyn, who submitted briefs containing invented cases, allegedly produced by generative AI tools. (LA Times)
- U.S. courts have documented 95 filings since June 2023 containing fictitious legal research, with judges often attributing these mistakes to uncritical use of AI-powered drafting. (Washington Post)
- High-profile penalties include $31,100 in sanctions for two U.S. law firms in 2025, and the High Court in London cited 18 fake case laws in a single dispute. (The Guardian)
The court's recent finding reframes the debate: while AI can generate plausible but false citations, human oversight remains the final line of defense. As George Cardona of the California State Bar stated, "Courts and clients must be able to trust that the filings attorneys submit are accurate, supported, and compliant with professional standards."
Dame Victoria Sharp, President of the King’s Bench Division, similarly warned, "Such tools can produce apparently coherent and plausible responses... but those... may turn out to be entirely incorrect."
The episode puts fresh emphasis on the need for robust quality control, with law firms urged to ensure both human and AI-generated content meet professional obligations before entering the record.
By the numbers:
- $31,100 — Sanctions levied on two U.S. law firms for AI-generated fake citations in 2025.
- 95 — Legal filings with fake research attributed to AI since June 2023 across the U.S.
- 18 — Fake case citations in a single UK case involving Qatar National Bank in 2025.