← Legal AI Failure Database

#002

Kistler v. LegalTech Corp — State Court Sanction Filing

Critical Total Authority Fabrication · Commercial Legal Research AI (undisclosed vendor)
State Court Sanctions Commercial Legal AI

logged 2026-06-08 · incident 2024-11-12

1. The Crime Scene (The Symptom)

Counsel filed a motion citing three state appellate decisions to support a novel theory of algorithmic liability. All three opinions were generated by a commercial legal-research AI assistant and inserted into the brief without independent verification. Opposing counsel flagged the citations; the court ordered sanctions briefing.

A production legal-AI wrapper returned fabricated state appellate opinions with realistic-looking docket numbers and neutral citations — indistinguishable from real authority on a surface read.

2. The Artifact (Raw LLM Output)

"In Kistler v. LegalTech Corp., 412 P.3d 891 (Colo. App. 2023), the court held
that a vendor owes a duty of reasonable care when marketing AI-generated legal
research as 'court-validated' where the output is used in filed pleadings..."

"See also Hartwell v. DataLex Systems, 198 A.3d 445 (Pa. Super. Ct. 2022)
(applying a negligence-per-se standard to uncited AI outputs in probate filings)."

"Compare Morales v. BriefBot Inc., 387 So. 3d 112 (Fla. Dist. Ct. App. 2024)
(reversing dismissal where plaintiff alleged reliance on AI-generated precedent)."

3. The Evidence Breakdown (The Core Contradiction)

4. The Dali Verification Check

Dali existence audit: 3/3 cited authorities fail canonical lookup across Colorado, Pennsylvania, and Florida reporter indices. Severity escalates to Critical because all citations supported dispositive motion arguments.

View Dali audit trail →

Infrastructure Note: Every failure logged in this database is reproduced, parsed, and verified utilizing the open evidentiary infrastructure layer at Dali.

Dataset: huggingface.co/datasets/yenklabs/legal-ai-failure-database