Skip to content

Risk intelligence, not legal advice

Track every AI hallucination sanction before it becomes your firm’s problem.

Editorial review

CONFIRMED

Every record on this site is reviewed by licensed counsel before publication and re-checked against its primary source on an ongoing basis.

Not legal advice. Nothing here creates an attorney-client relationship or substitutes for a licensed attorney reviewing your specific matter.

Read our editorial standards →

Lex Machina Review logs documented AI citation-hallucination and sanction rulings worldwide as they happen, each record sourced to the primary court order, alongside an obligations tracker and tool accuracy benchmarks. Freshness and per-record sourcing are the whole point.

Cases tracked here
1,382
Ruling cadence
New rulings logged as verified

Choose a task

Flagship tracker

Latest from the Risk Digest

Most recently updated records, with verification dates shown when available.

The rest of the dossier

Every case cross-links to the workflow that would have prevented it, the rule it turned on, and the tool it implicates.

Workflows

Separate charges from convictions in Chinese espionage cases

Public case lists of Chinese espionage prosecutions mix charges, pleas, and convictions. This explainer shows how to separate proven outcomes from announced ones, why DOJ press-release inventories are biased toward high-profile cases, and why any 2024-2026 list must be treated as unverified until checked against court dockets.

For attorney

Browse Workflows

Regulation

Which Federal Campaign Advertising Rules Are Verified?

A verified Q3 2026 baseline for campaign advertising by or on behalf of federal candidates, grounded in the four 11 CFR 110.11 disclaimer requirements — payor identification, clear-and-conspicuous presentation, broadcast audio and candidate-approval statements, and small-item exemptions. Each claimed obligation is flagged confirmed or unverified so counsel can separate settled rules from unsupported claims about enforcement history and AI-disclosure duties.

Scope US federal

Browse Regulation

Evaluations

Is open-source Muse Glimmer safe for legal work?

Meta's open-source Muse Glimmer 30B is its most permissive model release, but permissive licensing does not make it practice-ready. This risk record weighs what Apache 2.0 actually permits against independent hallucination-risk signals, the absence of any public legal-benchmark result, and the ethics duties that bind lawyers regardless of deployment.

Tool Muse Glimmer 30B

Browse Evaluations

Sourced, reviewed, and re-checked

Claims on this site are tied to primary sources — court orders, docket entries, ABA and state bar opinions, EU AI Act text, and named benchmark studies — never to derived or unattributed statistics. A licensed-attorney reviewer confirms new records before publication, and every record distinguishes a court-confirmed outcome from a merely reported one.

Read the full editorial standards →