NetGoodIndexSubmit a correction

Harm Ledger · verified · Computer Science

Lawyers sanctioned after ChatGPT fabricates cases in Mata v. Avianca

On 22 June 2023 the S.D.N.Y. sanctioned attorneys $5,000 for submitting nonexistent opinions generated by ChatGPT in Roberto Mata’s personal-injury suit against Avianca.

22 Jun 2023Tier 2 Significant HarmMethodology 0.1

Current score

−0.21

3 base · Significant Harm (tier 2 of 5, 3 pts)
× 0.3500 attribution · Material acceleration
× 1.0000 evidence · Overwhelming validation or measured outcome
× 1.0000 realization · Realized outcome
× 0.4000 durability
Event-level product before credit split: 0.42

Documented court harm of limited scale (tier 2). Modest model attribution because lawyers had a duty to check. Court record is conclusive evidence and realization. Precedent is lasting; the $5,000 sanction is small.

What happened

Judge P. Kevin Castel found that counsel submitted fake citations and then doubled down after the court asked for copies. The order became the template warning that lawyers must verify AI research. Harm is to the integrity of a federal proceeding and to the client’s representation, not a large population injury.

Model attribution

−0.21

GPT

Generated fabricated case names, quotes, and fake opinions used in filings.

ChatGPT produced the false authorities; attorneys filed them without verification.

Attribution 0.3500 · Credit share 50% · OpenAI

Claims

  • ChatGPT generated nonexistent judicial opinions that were filed in federal court, leading to Rule 11 sanctions.

    outcome · supported

Sources

primary sources

Revision history

  • 13 Sep 2026 · 0.00 → 0.21

    Initial adjudicated seed score under methodology 0.1.