The Parade of Horribles: Our Study on How AI Screwed Lawyers Over

We ran the study. We read the sanctions orders. The record since 2023 is a graveyard of legal careers, and every headstone reads the same: a lawyer trusted an AI, the AI invented case law, and a federal judge found out.

The documented body count

• Mata v. Avianca (S.D.N.Y. 2023) — six invented cases, $5,000, the case that started it all • California (2025) — 21 of 23 citations fake, $10,000 • London High Court (2025) — 18 of 45 fabricated in an £89M case; contempt warning • Oregon federal court (2026) — ~$109,700 combined; the court called the national record a “parade of horribles” • Mississippi (2026) — two attorneys barred for two years • 9th Circuit (2026) — fines plus six-month suspensions • A public database now tracks 120+ hallucination cases and climbing

The trend line is the story: $5,000 fines in 2023 became six figures and disbarments by 2026. Courts now sanction lawyers for failing to catch the OTHER side's fake citations. Every sanctioned lawyer pled the same defense — “I didn't know the AI would lie” — and every court rejected it, because the duty of candor cannot be outsourced.

Why it will never happen here

This platform runs the Quadruple Check: hard-coded rules that forbid invented authority; a hand-verified library of 200+ landmark cases; an automatic audit that extracts every citation from every generated document and flags anything unconfirmed — Verified, Verify Manually, or Fabrication Risk; and you, the human, as the final check. We tested our auditor against the actual fabricated case from the Avianca scandal. It refused to bless it on sight.

Read the full study at /parade-of-horribles. Read our covenant at /quadruple-check. The argument should win or lose on its merits — never on its citations.