AI hallucinations in US courts increase nearly sevenfold
AI-generated fake case citations in US courts surged to 1,667 by mid‑2026, nearly seven times the 2025 count.Summary
A study found that by mid‑2026, attorneys cited AI‑generated hallucinated cases in 1,667 court matters, up from about 230 the previous year—a nearly sevenfold increase. The rise reflects broader adoption of legal‑research AI tools and their persistent error rates, with leading platforms such as Lexis+ AI and Thomson Reuters producing incorrect information 17% to over 34% of the time in benchmarks. Real‑world consequences include lawyer sanctions, disqualifications, and high‑profile admissions of fake filings by firms like Sullivan & Cromwell. Judges now treat such errors as serious misconduct, mandating human verification of every AI‑generated citation and calling for structured training for paralegals and associates. The article warns that without rigorous verification habits, generative AI remains a liability rather than a productivity aid for legal work.
(Source：Complete Ai Training)
