Mata v. Avianca: The AI Legal Case Study Every Lawyer Should Know

Mata v. Avianca, decided in 2023 by Judge P. Kevin Castel of the Southern District of New York, is the legal AI case study every attorney should know. It established standards for AI use in legal practice that resonate through ABA Formal Opinion 512 and subsequent state bar guidance.

The facts

Roberto Mata sued Avianca Airlines after a personal injury claim involving a metal serving cart on a flight. His attorneys, Steven Schwartz and Peter LoDuca of Levidow, Levidow & Oberman, opposed Avianca's motion to dismiss with a brief that cited multiple federal and state cases supporting their position.

The brief was unusual in one respect: many of the cited cases didn't exist. They were generated by ChatGPT, used by Schwartz for legal research, without verification.

What the court found

Six fictional cases were cited in the brief:

• Varghese v. China Southern Airlines Co Ltd
• Shaboon v. Egypt Air
• Petersen v. Iran Air
• Martinez v. Delta Airlines
• Estate of Durden v. KLM
• Miller v. United Airlines

None of these cases exist in any legal database. ChatGPT had generated them.

When opposing counsel and the court searched for the cases, they couldn't find them. The lawyers were ordered to produce the cases or explain the discrepancy.

The lawyers' initial responses compounded the problem:

• They submitted copies of "case excerpts" that were also AI-generated
• They claimed the cases existed in "online legal databases"
• They asserted the citations were accurate

When the court continued probing, the lawyers eventually admitted they'd used ChatGPT for research without verification.

The sanctions

Judge Castel imposed sanctions on the lawyers and the firm:

• $5,000 fine
• Required notification to the affected judges of the fictional cases
• Required notification to the clients
• Public reputational consequences

The sanctions were modest in financial terms. The reputational and professional consequences were severe.

Why it matters

Mata v. Avianca became the foundational AI legal cautionary tale because:

• It crystallized AI risks for the bar. Lawyers across the country read about it and adjusted practice.
• It established verification standards. ABA Formal Opinion 512 and state bar guidance reference the case implicitly or explicitly.
• It provided a clear example of what not to do. The sequence of errors (using AI, not verifying, doubling down when challenged) became a training case.
• It influenced court rules. Some courts now require disclosure of AI use in filings.

The cascading errors

The case study power is in the cascade:

Error 1: Used ChatGPT for legal research without understanding its limitations.

Error 2: Filed brief without verifying citations.

Error 3: When challenged, doubled down on the citations rather than investigating.

Error 4: Submitted "case excerpts" that were also AI-generated to "prove" the cases existed.

Error 5: Continued asserting citations were accurate even as evidence mounted that they weren't.

Each error compounded the previous. The original AI hallucination became manageable; the cover-up and doubling down made it career-defining.

The lessons for attorneys

The lessons are clear and have been repeated throughout this guide series:

• Verify every AI citation. No exceptions. Pull the case, read it, confirm the proposition.
• Understand AI hallucination. It's a feature of how language models work, not a bug to be fixed.
• Use enterprise-tier tools. Free ChatGPT consumer tier is not appropriate for legal research.
• Train attorneys and staff. AI competence (ABA Rule 1.1) requires education.
• Build verification into workflow. Don't rely on individual attorney discipline alone.
• When errors emerge, address them immediately. Don't double down. Don't cover up.

The ABA Formal Opinion 512 connection

ABA Formal Opinion 512 (July 2024) implicitly addresses Mata v. Avianca:

• Lawyers must understand AI capabilities and limitations (Rule 1.1 competence)
• Verification of AI output is part of due diligence
• Candor to tribunal applies to AI-generated filings (Rule 3.3)
• Supervisory obligations apply to AI use (Rule 5.1/5.3)

The Mata v. Avianca facts illustrate what happens when these requirements aren't met.

Subsequent cases

Mata v. Avianca wasn't a one-off. Subsequent cases have featured AI-generated fictional citations:

• Multiple federal cases with sanctions
• State court cases with disciplinary referrals
• Bar discipline cases involving AI misuse

The pattern is consistent: AI generates fictional citations, lawyer doesn't verify, court catches it, sanctions follow.

What changed in legal practice

After Mata v. Avianca:

• Major firms accelerated AI policy development
• Bar associations issued specific AI guidance
• Some courts began requiring AI disclosure in filings
• Legal AI tools added verification features and warnings
• Attorney training on AI ethics became standard

The case fundamentally changed how the legal profession approaches AI.

The cultural impact

Beyond the formal legal consequences, Mata v. Avianca became cultural shorthand for AI misuse:

• "Don't Mata v. Avianca yourself" — used in firm training
• Case studies in law school ethics courses
• Reference in CLE programs
• Citation in firm AI policies

The case is taught in law schools, included in ABA continuing legal education, and referenced in firm AI training programs throughout the U.S.

What we tell attorneys

Every attorney deploying AI should:

• Read the case opinion
• Understand how the errors cascaded
• Build verification into every AI-assisted workflow
• Train staff on the case as a cautionary example
• Maintain a culture where admitting mistakes is safer than covering them up

The case is a gift to the profession — a clear, public example of what not to do.

Bottom line

Mata v. Avianca is the most important case study in legal AI ethics. The facts are simple. The lessons are clear. The consequences for ignoring them are severe.

Every attorney using AI for any legal work should know this case, understand its lessons, and build verification discipline that prevents the next Mata v. Avianca.

The case isn't a barrier to AI use — it's the operating manual. Use AI; verify everything. Don't double down on errors. Be candid with the tribunal. These are practice fundamentals AI doesn't change.