Security Architecture
Security is ContractCheck's first principle. This page explains what we do — and what we refuse to do.
Never used for training
Your contract is covered by Anthropic's enterprise data policy — never used to train models, and auto-deleted within 30 days. Zero Retention available on request for enterprise customers.
Self-destruct mode (Pro)
When enabled, contract and report are hard-deleted 10 minutes after you close the tab. We can't recover them either.
Database-level isolation (RLS)
Supabase Row Level Security enforces per-user isolation at the database layer. Not application logic. Database-native.
End-to-end encryption
TLS 1.3 in transit. AES-256 at rest. pgcrypto on sensitive columns as a second layer.
Engineered to be unreadable
We have no admin endpoint that can read contract plaintext. It's architectural, not a promise.
No third parties
No Google Vision, AWS Textract, or third-party OCR or analytics. Contracts only ever touch Anthropic.
For full technical detail, see SECURITY.md in our public GitHub repo.