Aegis helps AI teams show what data they used to train or fine-tune a model — without exposing that data — and turns that into evidence they can share in docs, model cards, and (if needed) legal discovery.
card-snippet - HF model card blocks
dataset-card - HF dataset cards
badges - Community or Notarized
Process your training data locally. Nothing leaves your machine.
Generate manifest, Bloom filter fingerprint, and reports.
Add to HuggingFace cards, documentation, or legal workflows.
# Install Aegis pip install aegis-client # Generate evidence for your training data aegis-cli attest --dataset ./training_data --out proof.json # Check content against training set aegis-cli verify --manifest proof.json --dataset "sample text" # Generate EU GPAI report aegis-cli report --manifest proof.json --out report.md # Check against opt-out registries aegis-cli optout-scan --dataset ./data --spawning-json optouts.json # Generate copyright policy aegis-cli policy --interactive --out COPYRIGHT_POLICY.md
Fine-tuning open or closed foundation models without a big legal team. Get evidence generation in minutes, not weeks.
Publishing on HuggingFace and want better cards and more reuse. Add training evidence directly to your model cards.
Selling models or AI apps in EU/UK/US and need to show you respected opt-outs and licensing.