Sample Certificate

This is what your auditor sees.

Every model trained with Paragon generates a Fairness Certificate automatically. Below is a sample certificate for a fictional company — Color Health — trained on 1000 Genomes data across 5 ancestries.

Open Full Certificate ↗ Try the Interactive Demo Generate Your Own

This sample certificate was generated by running:


paragon train --input data.csv --sensitive-column ancestry --target-column disease_risk --d 32

What's Inside

Every certificate includes:

Metrics

Task performance (R², AUC, accuracy)
Per-group breakdown across all demographics
Equity ratio (worst / best group)
Sensitive attribute leakage score

Compliance

EU AI Act Article 10 assessment
EU AI Act Article 15 assessment
Bottleneck dimension (the fairness parameter)
Dataset size and group counts

Integrity

SHA-256 model hash (tamper detection)
Certificate hash for verification
Signed by Paragon Verification Authority
Verifiable at paragondao.org/verify

Three Formats

One training run. Three outputs.

JSON

Machine-Readable

For CI/CD pipelines, automated audits, and programmatic verification.

HTML

Visual Certificate

For auditors, board presentations, and compliance documentation. Prints to PDF.

Badge

Embeddable

For your website and product pages. "Fairness Certified by Paragon."