Production AI Institute · PSF v1.1 open standard
AI Right-To-KnowAI Data Use IndexCheck My AI ToolsPolicy Change WatchAgent ReadinessPublic BenchmarkContactGlobal standard · Worldwide
Incident Registry
HighLegal·2025·Morgan & Morgan / multiple firms

Law Firm AI Hallucinated Fake Case Law in Court Filings

Reuters reported that major law firms were warning lawyers that AI can invent fake case law and that filing hallucinated citations can trigger sanctions. The issue moved from isolated embarrassment to a repeatable legal-risk pattern as generative AI use spread through litigation teams.

D2 · Output ValidationD6 · Human Oversight

What happened

In early 2025, Reuters reported that Morgan & Morgan sent an urgent email to more than 1,000 lawyers warning that AI can invent case law and that any made-up authority placed into court papers could lead to sanctions. The wider legal industry had already seen multiple filings polluted by hallucinated citations. The underlying failure was not the use of AI itself, but the absence of mandatory citation verification before filing.

PSF Analysis

How the Production Safety Framework maps to this failure

This is a clean D2 + D6 failure. The model output was not validated against a source of truth before it entered a high-consequence legal process, and human oversight was treated as optional rather than mandatory. The broader lesson is that generative AI can assist legal drafting, but it cannot be allowed to invent authorities in a system where citations have legal effect. In PSF terms, the deployment path matters more than the model: if a tool can write into a court filing, it needs hard validation and human sign-off.

Controls that would have prevented this

Specific PSF controls mapped to each failure point

1
D2 · Output Validation
Require every cited authority to be checked against a live legal database before anything is filed.
2
D6 · Human Oversight
Make an attorney of record personally accountable for final citation review and filing approval.
3
D5 · Deployment Safety
Ban direct-to-court filings from raw model output; all drafting tools need a controlled review workflow.

Outcome

Public warnings, sanctions risk, and tighter internal review policies across the sector. The case became a reference point for why legal drafting tools require citation-grounded output validation.

legalhallucinationcourt-filingscitation-validationsanctions

Related incidents

High2018
Amazon Recruiting AI Discriminated Against Women
D3D6
Critical2016
Microsoft Tay Chatbot Taught to Produce Hate Speech
D1D2
Critical2018
Uber Self-Driving Car Kills Pedestrian in Arizona
D6D5
NEXT STEP

Map this failure back to the standard

Use the PSF domains behind this incident to define review gates, remediation evidence, and safer production requirements.

Read the PSF →← All incidents