What Mata v. Avianca actually said about lawyers using AI

Mata v. Avianca is the case lawyers cite as proof that using AI is too dangerous to touch. The opinion itself says close to the opposite. The two New York lawyers in Mata were not sanctioned for using ChatGPT. They were sanctioned for what they did after it handed them six cases that did not exist.

What did the court actually say about using AI?

Judge P. Kevin Castel decided the case in the Southern District of New York in 2023. The first paragraph of his sanctions opinion is the part almost nobody quotes: “Technological advances are commonplace and there is nothing inherently improper about using a reliable artificial intelligence tool for assistance. But existing rules impose a gatekeeping role on attorneys to ensure the accuracy of their filings.” The court did not ban AI. It restated a duty older than AI: you are responsible for what you put your name on.

If the problem was not ChatGPT, what was it?

Filing fake cases, and then defending them. The court found that the lawyers “abandoned their responsibilities when they submitted non-existent judicial opinions with fake quotes and citations created by the artificial intelligence tool ChatGPT, then continued to stand by the fake opinions after judicial orders called their existence into question.”

The court spelled out the damage. Opposing counsel spent time and money exposing the fakes. The court's own time was pulled from other cases. Real judges had bogus opinions attributed to their names. The six fabricated cases even had names of their own, Varghese, Shaboon, Petersen, Martinez, Durden, and Miller, and not one of them existed.

Castel was explicit that the turning point was the response, not the research. Had the lawyers come clean when the cases were first questioned, he wrote, the record “now would look quite different. Instead, the individual Respondents doubled down.” The sanction that followed, a $5,000 penalty and an order to mail letters to each real judge whose name had been attached to a fake opinion, was for that, not for opening ChatGPT.

What duty does Mata leave for your firm?

One duty, and it is not new: a human confirms that what goes out is real before it goes out. A citation from an AI tool gets the same check a citation from a first-year associate gets. In July 2024 the American Bar Association made the AI version explicit in Formal Opinion 512, its first formal ethics opinion on generative AI. Submitting a tool's output without independent review, it warns, “could violate the duty to provide competent representation” under Model Rule 1.1. The tool does not hold that duty. The lawyer does.

Does this only matter for litigators?

The fact pattern in Mata was a court filing, but the principle is not limited to briefs. Anything that leaves the firm as your work product, a client letter, a contract, a research memo, carries your name and your responsibility for its accuracy. The lesson is not to keep AI out of the practice. It is to put a verification step between the draft and the door.

That step is a process decision, made once and built into the workflow, not a judgment call made nervously on every document. Designing it into the tools a firm already uses, so the speed of AI does not cost the firm the accuracy it is on the hook for, is the kind of work JurisLabs does. The first call costs nothing.