Skip to content
Submit incident
AII-1001-94181
ConfirmedA court, a regulator or the organization itself has stated the core facts, or the record cites a court or regulator source.

Anthropic's Claude Generates Fake Citation in Copyright Lawsuit

Anthropic admitted in a copyright lawsuit that its AI chatbot Claude fabricated a source citation, including an inaccurate article title and false author names, which appeared in expert witness testimony. The company acknowledged the error as an "honest citation mistake" and formally apologized to the court.

13 May 2025 Β· Anthropic PBC Β· Reliability and hallucination
13 May 2025
AII-1001-941813 sources ↓
Quick facts
Category
Reliability and hallucination
Country
United States
Source
Reuters
Date
13 May 2025
Drafted by
Raidu Incident Detection

What happened

Anthropic's attorneys asked Claude to generate a legal citation for an academic article titled 'Binomial Confidence Intervals for Rare Events: Importance of Defining Margin of Error Relative to Magnitude of Proportion' by Owen McGrath and Kevin Burke. Although Claude correctly identified the publication title, year, and link, it generated an inaccurate article title and invented false author names.

The fabricated citation appeared in a declaration by Anthropic employee Olivia Chen, who served as an expert witness in a copyright lawsuit brought by music publishers including Universal Music Group. The error passed through manual review and was later discovered by opposing counsel during court proceedings, prompting Judge Susan van Keulen to demand an explanation from Anthropic.

Reported impact

Music publishers accused Anthropic's expert witness of violating court rules by using Claude to develop arguments and citations in her declaration. Anthropic acknowledged the citation error, describing it as an 'honest citation mistake and not a fabrication of authority,' though the underlying article does exist and supports Chen's statement.

Index analysis

Claude's error involved hallucinating author names and an article title while retaining correct metadata about the publication, year, and URL. Anthropic stated that additional citation and wording mistakes from Claude also appeared in the footnotes during the formatting process, though the company did not release a complete list of faulty citations.

Governance controls

Anthropic's manual review process failed to detect the fabricated citation and other errors before the declaration was filed with the court. Anthropic's attorney was required to formally apologize to the court for the errors.

Classification

Category
Reliability and hallucination
Industry
Arts, entertainment and recreation
Country
United States
Harm status
Actual harm
Intent
Unintentional
Timing
Post-deployment
Harm types
Reputational
Harmed
Business
Failure modes
Hallucination, Fabricated citation, Agent misaction
Consequences
Litigation, Public apology
AI systems
Claude

Sources and evidence

This record was researched and written by the Index, based on the following original reporting.

Reuters
Anthropic expert accused of using AI-fabricated source in copyright case
13 May 2025
news.bloomberglaw.com
Anthropic Expert Accused of Citing Fake Article in AI Lawsuit
13 May 2025
the-decoder.com
Anthropic's Claude Generates Fake Citation in Copyright Lawsuit