Two Law Firms Were Fined $31,000 After AI Invented the Cases Their Brief Cited
What happened
A California judge fined two U.S. law firms a combined USD 31,000 in May 2025 after discovering that a court brief they submitted contained multiple legal citations and quotations that did not exist. The cases cited were fabricated. The quoted text attributed to those cases was fabricated. The AI tools that generated the brief had produced confident, specific, entirely fictional legal authority, and the attorneys who filed it had not checked.
The brief was submitted in a civil case against an insurer. The law firms used Gemini, developed by Google, and Westlaw Precision with CoCounsel, a legal research tool built by Thomson Reuters, to conduct their research. Neither tool flagged the citations as nonexistent before filing. More to the point, no attorney in either firm confirmed the cited cases were real before the brief reached the judge. The California court issued both the fine and a public reprimand, marking the firms as part of a growing list of legal professionals sanctioned for the same class of failure.
This is not a rare event. Related incidents in the record include a psychologist who submitted legal filings with fictitious AI citations, a Walmart product liability case that cited invented cases, and a widely reported 2023 episode in which lawyers were sanctioned after submitting AI-generated case citations to a federal court. In each instance, a language model produced output that looked like legal research and functioned as legal research in the filing, but had no connection to any actual court decision.
For clients whose cases depend on these briefs, the risk is direct: a motion built on nonexistent authority can be struck, weakening or losing a case that might otherwise have held. For the legal profession, the fines and public reprimands are significant deterrents, but they arrive after the damage is done. The firms that paid them had already filed the brief, already submitted fabricated legal authority to a court of record, and already put their clients in jeopardy.
The underlying gap here is a verification problem with no current enforcement structure around it. A lawyer who uses an AI research tool has no reliable way to prove, after the fact, which queries produced which citations, whether the tool was asked to confirm the citations existed, or where the fabricated text entered the document. A provable record of what a system generated, what a professional reviewed, and what was actually filed would make that chain of custody visible. Without it, the profession is left relying on attorneys to manually verify every output from tools that are specifically marketed to reduce that kind of manual work.
Reported impact
- Affected parties
- Not publicly disclosed
- Harm type
- Not publicly disclosed
- Scale
- Not publicly disclosed
- Financial impact
- Not publicly disclosed
- Regulatory action
- Not publicly disclosed
Classification
Relevant governance controls
Governance control mapping is not available for this record.
- No controls mapped
Not publicly disclosed
Control mapping is analytical. It does not state that any control would have prevented the incident.
Sources and evidence
This record was researched and written by the Index. The event is also catalogued in the following database, which is listed for cross-reference.