The judgments that do not exist

In June 2023 a New York federal judge, P. Kevin Castel, sanctioned two lawyers and their firm with a $5,000 penalty. They had filed a brief with non-existent judicial opinions, fake quotes and citations created by the tool ChatGPT. The case is Mata v. Avianca.

In the opinion the judge recalls that existing rules impose a gatekeeping role on attorneys to ensure the accuracy of their filings. The problem was not using a new tool. It was not checking.

Not an isolated case

In June 2025 the High Court of England and Wales returned to the issue in two cases decided together, Ayinde and Al-Haroun. The judgment was given by Dame Victoria Sharp, President of the King's Bench Division, and Mr Justice Johnson.

The principle is sharp. Those who use these tools for legal research have a professional duty to check the accuracy of that research against authoritative sources before using it in their work. And the duty also applies to those who rely on the work of others who used them.

Meanwhile Damien Charlotin maintains a database of decisions in which a court found, or implied, that a party relied on invented content. It does not collect mere suspicions. It collects decisions.

The real issue

One could read all this as an argument against the tools. I read it differently.

An invented citation reaches the court when a chain breaks: the drafter does not verify, the signatory does not reread, whoever organises the work does not plan a check. It is a process failure, not a technology failure.

That is why the answer cannot be only disciplinary. Work has to be designed. Every citation needs a source that someone has opened. Every filing needs a person who answers for it. And verification must be a visible step, not an act of faith.

The professional of the future is not the one who uses the tools most. It is the one who can show what they checked.