On 22 June 2023, a federal judge in New York sanctioned two lawyers and their firm for filing a brief that cited cases that did not exist. They had found the cases through ChatGPT. The judge did not ban AI use. He said lawyers must still verify what they file.
What happened
The case is Mata v. Avianca, in the Southern District of New York, decided by Judge P. Kevin Castel. According to LawNext's account of the order, attorneys Peter LoDuca and Steven A. Schwartz, with their firm Levidow, Levidow and Oberman, submitted a brief citing fabricated opinions with invented quotes and citations.
The court ordered:
- a $5,000 monetary penalty on the lawyers and the firm
- notice to their client about the sanctions
- notice to each judge whose name was falsely attached to the fabricated opinions
What the judge said about AI
The reported reasoning is narrow. Using AI is not inherently improper, but existing rules make attorneys the gatekeepers who verify their filings. The court found that the attorneys abandoned that duty.
The reported timeline is also important. The lawyers kept defending the citations after the court's orders questioned whether the cases existed, and did not disclose the truth until 25 May, after an order to show cause was issued. The judge suggested sanctions might have been avoided had they come clean sooner. The court found bad faith based on "conscious avoidance" and false or misleading statements to the court.
So two failures stack: relying on unverified output, and then not correcting the record.
Lessons that apply beyond law
The profession differs from yours, but the pattern is general.
- The user owns the output. The tool's authorship did not shift responsibility.
- Fluent is not true. Invented citations looked normal because the format was right.
- Delay makes it worse. The recorded penalty followed the cover up as much as the original mistake.
If your team uses an AI tool for anything someone else relies on, such as code, reports or customer answers, name a person who checks it. See a verification checklist for AI generated references.
A note on tone
It is easy to read this case as a story about careless people. The order, as reported, is more useful as a story about process. A single check in a primary database would have caught the problem early. A prompt reply to the first doubt raised would likely have reduced the outcome. Both steps are cheap compared with the cost of the sanction and the notice to every named judge.
What we did not verify
The court order itself could not be retrieved for this post, so the details above come from LawNext's report of it. Sources differ on the exact number of fabricated cases, so this post does not give one. Read the opinion for exact language.
Quick check
Did the court say lawyers cannot use AI?
No. As reported, it said using AI is not inherently improper, but attorneys must verify their filings.
Why was the penalty tied to the aftermath?
The reported reasoning points to the lawyers continuing to defend the citations after the court questioned them, which supported a finding of bad faith.
Sources
Last reviewed: 19 October 2026.
