Did you know that lawyers cited decisions that did not exist? In the Mata v. Avianca case, in two thousand twenty-three, references produced by ChatGPT are filed in court. A text presented with confidence can look like a source without being a judicial decision. The problem is not just trying a tool. References and citations were not checked, and then continued to be defended despite the questions. Our abstract document shows the context, without in turn fabricating a quote attributed to a judge.
The order of June 22 provides for a penalty of five thousand dollars, due jointly by the lawyers and their firm. It’s not five thousand dollars each, nor is it a criminal conviction of a robot. It examines the behavior and presentation of false documents here.
A reference must lead to a decision that can be found and read. The name of a case and a compelling quote are not enough. AI helps write, but this story shows the cost of defending a non-existent source as an established fact.
What we know
The order of June 22, 2023 imposes $5,000 of joint sanction on lawyers and the firm, not $5,000 each. The court examines human acts, not the criminal culpability of a robot.
To go to the source
References consulted on October 12, 2026.
Signaler une erreur ou apporter une précision
