AI Court Sanctions Database: Documented Cases of AI-Generated Citation Failures
A working list of federal and state court cases where attorneys filed AI-generated citations and faced sanctions, show-cause orders, or grievance referrals. Each case is tagged by failure mode and linked to the official docket plus a Veritas case analysis.
This database is maintained by Veritas, a pre-filing citation verification platform built after the canonical Mata v. Avianca sanctions order. New cases are added as published orders surface. The dataset is free to read, cite, and link to. CC BY-NC 4.0.
Failure modes
Every documented AI-sanctions case to date fits into one of the following operational failure modes. The taxonomy is the working shape of the underlying problem.
- Citations Not Located in Reporter. Citations to cases that do not exist in the named reporter. The dominant failure mode in every documented AI-sanctions case to date.
- Internal Review Failure. The firm had a review process. The process did not catch the AI-generated cite before filing. A signal that firm-level controls have not adapted to generative drafting.
- Pro Se Failure. Self-represented attorney or unsupervised solo practitioner filed AI-generated work without an internal verification step.
Citations Not Located in Reporter
Citations to cases that do not exist in the named reporter. The dominant failure mode in every documented AI-sanctions case to date.
Mata v. Avianca, Inc.
Six fabricated citations produced by ChatGPT cited in opposition to a motion to dismiss. Monetary Rule 11 sanctions on two attorneys and the firm; required notification of the judges falsely identified as authors.
Park v. Kim
Pro se attorney filed a reply brief with citations generated by ChatGPT that did not exist. The Second Circuit referred her to the court's grievance panel.
Morgan & Morgan (Wyoming Class Action)
BigLaw plaintiffs' firm filed motions in limine containing AI-generated citations that did not exist. The court ordered show cause; the firm's internal review process had not caught the cites before filing.
Internal Review Failure
The firm had a review process. The process did not catch the AI-generated cite before filing. A signal that firm-level controls have not adapted to generative drafting.
Morgan & Morgan (Wyoming Class Action)
BigLaw plaintiffs' firm filed motions in limine containing AI-generated citations that did not exist. The court ordered show cause; the firm's internal review process had not caught the cites before filing.
Pro Se Failure
Self-represented attorney or unsupervised solo practitioner filed AI-generated work without an internal verification step.
Park v. Kim
Pro se attorney filed a reply brief with citations generated by ChatGPT that did not exist. The Second Circuit referred her to the court's grievance panel.
How this database is maintained
Each entry references a published court document. Reporter citations are read in their named volumes; quotations from orders are checked against the docket. Cases are added when a published sanctions order, show-cause order, or appellate referral enters the federal or state record. Submit a case Veritas should track by emailing the docket link to hello@veritaslaw.app.
Why every case failed the same test, in one paragraph
Every case in this database failed the same operational test: the cite was not checked against the reporter before filing. Veritas runs the audit layer that returns a verdict on every citation in a draft brief (located, partial, not located in reporter) and produces a Verification Certificate for the matter file. Do not sign a brief that has not been verified.