Articles 1–4 told you how the council thinks. Article 13 answers the question every certification authority should be forced to answer: who watches the watchdog? The Charter's answer: everyone.
"Trust through transparency. If we govern AI on behalf of humanity, humanity must see everything we do."
The Watchdog platform is the council's own glass house: a real-time database of every certified system (status: active, suspended, revoked — with safety-case summaries and incident history) · live council operations (Byzantine Council voting statistics, meeting minutes) · public incident disclosure (what failed, what the investigation found, what was corrected) · and a public participation portal where anyone can raise a concern about a certified system.
Every legacy compliance vendor asks you to trust their letterhead. The Charter inverts it: a certificate is only as credible as the certifier's own transparency. That's why every Watchdog Certificate carries a public verify URL — and why the certifier's decisions, statistics and failures are published too. Power without transparency becomes tyranny; certification without transparency becomes marketing.
Watchdog Certification is the operating revenue of the council — independent assessment against 20+ frameworks, an Ed25519-signed certificate, and a verification link your regulator clicks. The incident layer ships today as agent-incident-reporter-mcp: hash-chained, signed, tamper-evident incident records mapped to EU AI Act Article 73. Get certified →
Care, not control (Art 1) → proof, not promises (Art 2) → never trust one monitor (Art 3) → never assume you know what humans want (Art 4) → and let the public see all of it (Art 13). Five articles in, the pattern is consistent: every power the council holds comes paired with a mechanism that checks it.