README.md

metadata

title: CROVIA · Evidence Machine
emoji: 🔬
colorFrom: blue
colorTo: indigo
sdk: gradio
sdk_version: 4.36.1
python_version: '3.11'
app_file: app.py
pinned: false
short_description: AI forensic evidence console with live data
tags:
  - audit
  - provenance
  - ai-act
  - compliance
  - forensics
  - cryptography
  - evidence

CROVIA · Evidence Machine

The world's first AI forensic evidence console.

Type any AI model. Get a complete, cryptographically anchored, temporally verified evidence package showing exactly what training data documentation exists — and what doesn't.

What this does

• 200+ models monitored with 20 NEC# documentation requirements each
• Pedersen commitments anchor every observation cryptographically
• Chain height 3,000+ — continuous, append-only temporal proof
• Regulatory mapping — jurisdictions affected by each documentation gap
• Citation engine — one-click, legal-grade evidence statements
• Outreach tracking — was the organization contacted? Did they respond?
• Peer comparison — how does this model compare to its organization and industry?

How to use

1. Type a model ID (e.g. google/gemma-3-12b-it) in the command bar
2. Click INVESTIGATE
3. Read the signal strip, NEC# constellation, forensic report, and citation

What you see

• Signal Strip — Trust level, evidence strength, chain height, NEC# gaps, jurisdictions
• NEC# Constellation — Interactive SVG showing all 20 documentation requirements as nodes
• NEC# Grid — Color-coded grid of present/absent/critical elements
• Forensic Report — Full monospace report with cryptographic anchors
• Jurisdictions — Which regulations apply to the observed gaps
• Citation Engine — Copy-paste evidence statement for legal or journalistic use

CEP Capsules tab

Backward-compatible inspector for reference CEP capsules from Crovia/cep-capsules.

Important

This is observation, not judgment. All data derived from publicly observable artifacts. No model audit. No legal claim. Presence/absence only.

Source: registry.croviatrust.com

License

Apache-2.0