A judge from a US district court has rescinded his earlier ruling in a securities case involving the biopharmaceutical company CorMedix after legal representatives pointed out inaccuracies in his opinion, including fabricated quotes and incorrect case references. These errors are similar to issues seen in other legal contexts attributed to the use of artificial intelligence.
In a letter addressed to Judge Julien Xavier Neals of New Jersey, lawyer Andrew Lichtman highlighted a “series of errors” in the judge’s refusal to dismiss a lawsuit brought against CorMedix. The claimed errors included misrepresentations of outcomes in three separate cases and multiple instances of fictitious quotes attributed to other judicial rulings.
According to a report from Bloomberg Law, a revised notice posted to the court record on Wednesday stated that the initial opinion and order were issued in error, with plans for a subsequent opinion and order to follow. It is not uncommon for courts to amend decisions after issuance for minor corrections, but significant revisions, such as altering key paragraphs or redacting extensive sections, are less frequent.
While there is currently no confirmation that AI tools were utilized in this case, the errors in citations exhibit characteristics often associated with AI-generated text. Such issues have been increasingly noted as legal professionals incorporate tools like ChatGPT into their research processes. For instance, attorneys representing MyPillow founder Mike Lindell faced penalties earlier this month for relying on AI-generated citations. Additionally, Anthropic attributed an “embarrassing” citation error to its Claude AI chatbot during its legal dispute with music publishers, among other similar incidents showcasing the limitations of AI tools in legal work.
