Anthropic has addressed claims regarding its use of an AI-generated source in an ongoing lawsuit with music publishers, clarifying that its Claude chatbot made an “honest citation mistake.”
The citation error was part of a legal filing submitted by Anthropic’s data scientist, Olivia Chen, on April 30, defending against allegations of unauthorized use of copyrighted lyrics to train the Claude model. During a court hearing, an attorney representing Universal Music Group, ABKCO, and Concord described the sources mentioned in Chen’s filing as a “complete fabrication,” suggesting they were the result of hallucinations from Anthropic’s AI.
In a filing submitted on Thursday, Anthropic defense attorney Ivana Dukanovic defended the authenticity of the contested source, stating that Claude was indeed utilized to format the legal citations. She noted that while incorrect volume and page numbers were identified and amended through manual checking, other wording inaccuracies remained unnoticed.
Dukanovic acknowledged the issue, saying, “unfortunately, although providing the correct publication title, publication year, and link to the provided source, the returned citation included an inaccurate title and incorrect authors,” emphasizing that the error did not constitute a “fabrication of authority.” The company expressed regret over the confusion and inaccuracy created by the citation error, describing it as “an embarrassing and unintentional mistake.”
This incident highlights the increasing challenges associated with using AI tools for legal citations. Recently, a Californian judge criticized two law firms for not disclosing the use of AI in crafting a supplemental brief filled with “bogus” materials that did not exist. Moreover, a misinformation expert revealed in December that ChatGPT had also produced fabricated citations in a legal filing he submitted.
