Open Source

HalluCase

The Legal AI Hallucination Registry

A living CVE-style registry for tracking, reproducing, and preventing hallucinations in legal AI outputs.

GitHub →npm →← All products

1,450+

Global hallucination incidents documented

$145K+

US sanctions in Q1 2026 alone

Hallucination type categories

By Q1 2026, over 1,450 court cases globally involved AI-fabricated legal citations, with five to six new documented incidents appearing every single day. HalluCase is the only vendor-neutral CVE-style registry built to systematically track these incidents.

Every report receives a unique HC-XXXXXX identifier with structured fields for hallucination type, severity level, domain, and AI model metadata. The registry ships with a REST API, CLI, and SQLite backend with FTS5 full-text search.

Key Features

CVE-style incident tracking with HC-XXXXXX identifiers and 9 hallucination classifications

REST API with full CRUD, search/filter, pagination, and schema introspection

CLI with search, submit, import (CSV/JSON), stats, and interactive report creation

5-level severity taxonomy (critical to info) and per-domain filtering

SQLite backend with FTS5 full-text search, UUID primary keys

Import pipeline from external sources including the Charlotin hallucination database

Loading…