The Council for the Safety of AI — the independent authority (CSOAI LTD, UK 16939677, founded by Nicholas Templeman) that sets AI-safety expectations, maintains crosswalks across 20+ governance frameworks, and issues Watchdog Certification.
An independent assessment of an AI system against applicable frameworks, issued as an Ed25519-signed certificate with a public verify URL. It proves identity and integrity — CSOAI issued this exact result and it hasn't been altered. It is not a legal guarantee of compliance.
Open the verify URL (csoai.org/verify/CERT-ID), or download the public key from /pubkey and verify the signature entirely offline — no contact with CSOAI required.
Article 50 transparency: new generative systems from 2 August 2026; legacy generative from 2 December 2026. Annex III high-risk from 2 December 2027. EU CRA full effect 11 December 2027.
Certification £199/mo (entry £29/mo) · bespoke 48h assessment £4,950 · Article 50 kit £999 · enterprise £1,499/mo. Every tier includes x402 pay-as-you-go per check.
Yes — agent card at /.well-known/agent.json (A2A), payable endpoint in acp.json (ACP + x402), remote MCP at /mcp/csoai for Claude and ChatGPT.
One legal entity, two brands. CSOAI certifies (csoai.org); MEOK AI Labs builds (meok.ai). proofof.ai and councilof.ai are built by MEOK, certified by CSOAI. The certifier is never the vendor.
Yes — the governance fabric is open source under github.com/CSOAI-ORG (290+ servers in the official MCP Registry, verified June 2026). Signed certificates are the commercial layer.