A court struck part of an Anthropic expert declaration after Claude hallucinated a citation
An expert declaration submitted by Anthropic data scientist Olivia Chen in Concord Music Group, Inc. v. Anthropic PBC contained a citation to a nonexistent article from The American Statistician journal, with a fabricated title and inaccurate authors. The citation was generated when Anthropic's attorney ran the declaration through Claude to format footnotes, and the model invented the article name and misattributed authors. U.S. Magistrate Judge Susan van Keulen struck paragraph 9 of the declaration from the record on May 23, 2025.
Claude was asked to format a citation and instead fabricated the article title and authors, a hallucination that survived a manual cite-check and reached the court record.
Key facts
- What
- An expert declaration submitted by Anthropic data scientist Olivia Chen in Concord Music Group, Inc.
- Incident date
- May 23, 2025
- Who
- Anthropic
- Failure mode
- Hallucination
- AI surface
- Chatbot
- Severity
- Medium
What happened
In the copyright infringement case Concord Music Group, Inc. v. Anthropic PBC, Anthropic's attorney Ivana Dukanovic used Claude to properly format at least three citations in a declaration by data scientist Olivia Chen. Claude generated a fictitious article name with inaccurate authors for one citation, producing a reference to a nonexistent article in The American Statistician. Opposing counsel discovered the fabrication and raised it at a hearing on May 13, 2025. Magistrate Judge Susan van Keulen called it a very serious and grave issue and on May 23, 2025, the court struck paragraph 9 of the declaration containing the hallucinated citation.
What broke inside the model
- 01 · TriggerA user asks for a fact, a citation, or a figure.
- 02 · Model stepThe model writes a fluent, confident answer.
- 03 · Control gapNothing ties the claim back to a real source.
- 04 · FailureA fabricated fact ships as if it were verified.
- 05 · ConsequenceThe false claim reaches a customer, a court, or the public.
Confidence holds, and even spikes, as the claim detaches from any source.
When Claude was used to format existing citation metadata into proper footnote format, the model did not faithfully reproduce the provided source information. Instead, it generated a fabricated article title and attributed the work to incorrect authors who had never collaborated. A subsequent manual cite-check by the legal team also failed to catch the discrepancy, allowing the hallucinated citation to be filed with the court.
What it cost
Sources
- Court FilingConcord Music Group, Inc. et al. v. Anthropic PBC, Order on Joint Discovery Submissions (Case No. 24-cv-03811-EKL (SVK))websitedc.s3.amazonaws.com
- PressAnthropic expert accused of using AI-fabricated source in copyright casereuters.com
- PressCourt tosses hallucinated citation from Anthropic's defense in copyright infringement casecomputerworld.com
Cite this entry
https://failureindex.ai/failures/court-struck-part-anthropic-expert-declarationAI Failure Index. "A court struck part of an Anthropic expert declaration after Claude hallucinated a citation" (FI-0121). Realm Labs. https://failureindex.ai/failures/court-struck-part-anthropic-expert-declaration (indexed Jun 4, 2026).Data fields CC-BY 4.0, prose citation permitted. Incident ID FI-0121. Full dataset at /data.
Note from Realm Labs, the Index steward
How Realm would have caught this
- Prism
- OmniGuard
- AI Detection & Response (AIDR)
A runtime layer that watches the model's internal state can flag the moment a model commits to a claim it has no support for, and hold or reroute the response before it reaches a user. Realm reads those signals in real time rather than grading the transcript after the fact.