Kansas Judge Demands Answers Over AI-Hallucinated Citations in Overstock Patent Case

A federal judge pressed Lexos Media IP's lawyers to explain AI-made citations in an Overstock patent case, with sanctions possible. They have until Jan. 5 to detail their roles.

Categorized in: AI News Legal
Published on: Dec 18, 2025
Kansas Judge Demands Answers Over AI-Hallucinated Citations in Overstock Patent Case

Judge orders patent attorneys to explain AI-created citations, sanctions on the table

On Dec. 17, 2025, Senior U.S. District Judge Julie A. Robinson (D. Kan.) ordered attorneys for Lexos Media IP to show cause why they shouldn't be sanctioned after briefs included nonexistent and incorrect legal citations generated by AI, according to Law360. The case is a patent infringement suit against Overstock.com Inc.

Counsel from Fisher, Patterson, Sayler & Smith; Buether Joe & Counselors; and Seth Law were given until Jan. 5 to detail their roles in drafting, reviewing, and filing the defective briefs. The court flagged "nonexistent quotations," "nonexistent and incorrect citations," and "misrepresentations about cited authority," per the report.

The judge noted attorney Sandeep Seth (Seth Law) had already filed a declaration acknowledging his role and stating the incorrect citations originated from AI research that was not independently verified.

What's at stake

This is a Rule 11 moment. Submitting invented quotes or phantom cites-AI-generated or not-risks sanctions, damages credibility with the court, and exposes clients to avoidable harm. Courts expect reasonable inquiry before filing. AI doesn't change that duty.

  • Rule 11 exposure: Filings must be grounded in law and fact after a reasonable inquiry. See FRCP 11.
  • Pattern recognition by courts: Judges have already sanctioned counsel for AI-fabricated citations, e.g., the Avianca matter in S.D.N.Y. (case docket).
  • Client risk: Bad citations can skew merits, inflate costs, and trigger fee-shifting or adverse orders.

Practical guardrails for legal teams

  • Ban unsupervised AI citations: No AI-generated cite or quote enters a filing without human verification and a citator check.
  • Read the source: Require attorneys to open and read every cited authority. No summaries. No screenshots. Confirm pincites and quotations.
  • Document the inquiry: Keep a short audit log per filing: who verified each cite, what sources were reviewed, and when.
  • Use a citator every time: Shepardize/KeyCite/BCite or equivalent. Note subsequent negative treatment in the record.
  • Separate drafting from research: If AI assists with drafting, require citations to be inserted only after independent research.
  • Policy and certification: Adopt an AI-use policy for filings. Consider an internal certification step when briefs go to partners for sign-off.
  • Vendor diligence: If using research tools with AI features, confirm how citations are generated and what guardrails exist.
  • Training and drills: Run team exercises spotting fake cites, verifying quotations, and tracing authorities back to the reporter.

Bottom line: AI can speed up early thinking, but it cannot replace your duty to verify. Courts aren't grading effort-they're grading accuracy.

If your team is formalizing AI practices, consider structured training focused on verification and prompt discipline. A practical starting point: prompt engineering resources.


Get Daily AI News

Your membership also unlocks:

700+ AI Courses
700+ Certifications
Personalized AI Learning Plan
6500+ AI Tools (no Ads)
Daily AI News by job industry (no Ads)
Advertisement
Stream Watch Guide