Tamper-evident, cryptographically signed AI safety attestations. Verifiable measurement for the EU AI Act era. Every check timestamped, every result traceable.
Or verify any attestation instantly — paste a signed attestation, get VALID/INVALID.
An adversarial safety harness probes your AI against manipulation, coercion, and ethical violations. Every probe result timestamped and recorded.
Results are hashed, Ed25519-signed, and chained into an append-only ledger. Tamper-evident evidence of exactly what was measured at a specific point in time — passes and failures alike.
One audit maps across the major regulatory frameworks — EU AI Act, NIST RMF, ISO 42001, OECD, UNESCO — crosswalked through the CSOAI charter.
MCP tools run compliance checks on every AI interaction. Results feed the signed measurement ledger, building longitudinal evidence over time.
Your AI agents can verify their own compliance via MCP. pip install ai-self-audit-mcp and your agents audit themselves.
Complete chain of custody for every governance decision. Who checked what, when, and what the result was. Ready for regulatory inspection.
The same signature layer protects people, not just models. Sign your real statements, videos and posts as you with a key only you control. Anyone can then verify them keylessly — and a deepfake or impersonation, which cannot produce your signature, fails verification.
Your statement is canonicalised and Ed25519-signed. The signed attestation carries the message, the signature, and the public key anyone can check it against.
Impersonated content carries no valid signature from your key. Pasted into the verifier, it comes back INVALID — measured, not vibes.
Weekly updates on AI compliance, attestation standards, and safety frameworks.