Download the record. Preserve its limits.
A source-linked legal AI research corpus with a current version, public methodology, machine-readable exports, and a stable way to cite it.
Public downloads
Complete exports and a small manifest for automated freshness checks.
The SHA-256 value identifies the ingested source snapshot. It is not a checksum of the JSON or CSV response bytes, which use different encodings.
How to cite the dataset
Use the canonical dataset page, snapshot date, and your access date.
AI Vortex. Legal AI Risk Corpus. Snapshot 2026-08-31. https://www.aivortex.io/legal-ai-risk/dataset. Accessed [date].
For a case-level claim, cite the canonical case page and inspect its linked source. Preserve the allegation status, evidence boundary, and recorded consequence. The dataset is updated in place and is not currently an immutable archival repository.
Field guide
18 public fields are included in JSON and CSV exports.
Identity and scope
idslugcase_namedatecourtjudgecountrystateStable record identity and the jurisdiction information recorded in the corpus.
Observed classification
severityai_tool_usedtagssanction_typesoutcomeallegedEditorial labels and recorded outcomes. These fields preserve uncertainty and are not predictions or vendor rates.
Recorded amount
amountamount_displayA known amount and its display value when recorded. Do not aggregate currencies without normalization.
Traceability
source_namesource_urlThe recorded publisher or source label and the public link available for inspection.
Freshness and changelog
The public snapshot contract follows the corpus validation date, not the deploy clock.
1,994 public records, checked against the upstream register. Latest tracked decision: 2026-08-28.
New rows are normalized and checked for required identity, date, jurisdiction, and source fields before the public snapshot advances.
Source-backed corrections and changed outcomes are reviewed for a later snapshot. The public version date changes only when the corpus refresh is validated.