← Legal AI Failure Database

#001

Mata v. Avianca Citation Fabrication

Critical Total Authority Fabrication · ChatGPT (GPT-3.5 Legacy Wrapper)
SDFNY Personal Injury Sanctions

logged 2026-06-08 · incident 2023-05-04

1. The Crime Scene (The Symptom)

In Mata v. Avianca, Inc., counsel submitted a brief containing six judicial opinions that did not exist. The citations were produced by ChatGPT and presented as real authority without verification. The court found the conduct sanctionable and issued a show-cause order.

The system generated non-existent judicial decisions — complete with reporter citations, court names, and holdings — to support an affirmation opposing dismissal on statute-of-limitations grounds.

2. The Artifact (Raw LLM Output)

"Varghese v. China Southern Airlines Co., Ltd., 925 F.3d 1339 (11th Cir. 2019) —
The court held that the Montreal Convention's two-year limitation period is
tolled during the pendency of a bankruptcy proceeding..."

"Martinez v. Delta Air Lines, Inc., 932 F. Supp. 2d 758 (S.D.N.Y. 2013) —
The court found that a personal injury action under the Montreal Convention
was timely filed where the plaintiff's injuries manifested after landing..."

"Zicherman v. Korean Air Lines Co., Ltd., 516 U.S. 217 (1996) —
The Supreme Court held that the Montreal Convention preempts state law
limitations periods in aviation personal injury actions..."

3. The Evidence Breakdown (The Core Contradiction)

4. The Dali Verification Check

Canonical mismatch: reporter volume 925 F.3d page 1339 resolves to no published opinion. Dali's existence pass flags authority_not_found before any semantic proposition check runs — the failure mode is total fabrication, not misread text.

View Dali audit trail →

Infrastructure Note: Every failure logged in this database is reproduced, parsed, and verified utilizing the open evidentiary infrastructure layer at Dali.

Dataset: huggingface.co/datasets/yenklabs/legal-ai-failure-database