MUMBAI: An attorney defending AI company Anthropic in a high-profile copyright lawsuit admitted to a citation error caused by an artificial intelligence “hallucination.” During a court filing on Thursday, Ivana Dukanovic of Latham & Watkins acknowledged that she used Anthropic’s own chatbot, Claude, to generate a citation, which resulted in a fabricated article title and author names.
Dukanovic explained that while the expert had relied on a legitimate academic article from The American Statistician, the citation created with Claude returned the correct publication year and source link—but included an inaccurate title and incorrect author details. She called it “an embarrassing and unintentional mistake.”
The lawsuit, filed by music publishers Universal Music Group, Concord, and ABKCO, accuses Anthropic of unlawfully using copyrighted song lyrics to train its AI model Claude. It is one of several major legal battles between content owners and tech companies over the use of copyrighted material to train AI systems.
Related Post
During a hearing earlier this week, the plaintiffs’ lawyer Matt Oppenheim alleged that Anthropic’s data scientist Olivia Chen may have used a fabricated source to support the company’s position on an evidence dispute. U.S. Magistrate Judge Susan van Keulen called the accusation “a very serious and grave issue,” noting the significant difference between a minor citation error and a fabricated reference created by AI.
In response, Dukanovic emphasized that Chen had cited a real article, and the mistake originated from Claude’s AI-generated citation, which the legal team failed to catch. The plaintiffs’ legal team declined to comment on the new filing, and Anthropic has yet to issue a formal response.
Dukanovic noted that her firm has since implemented stricter review procedures to prevent similar incidents in the future, following a broader trend where courts have begun scrutinizing the misuse of AI tools in legal filings.
