AI Transparency Statement

Version 1 August 2026 · effective and last updated 1 August 2026 · last reviewed 1 August 2026

Purpose

Sacred Logic certifies other organisations' AI systems against the Human Values Framework. This page is about a narrower thing: the AI Sacred Logic itself uses on sacredlogic.org, how the self-assessment score is reached, and where a human is involved before anything is decided.In short: where AI is used — the chat assistant only. Where it is not used — the self-assessment score, and every certification decision. Human judgement is mandatory before any Charter Mark is issued, deferred, or declined, and, as stated below, no certification decision is ever made solely by AI.

The self-assessment score is not AI

The free self-assessment's EU AI Act risk classification and Human Values Framework readiness score are produced by fixed, rule-based logic — a short decision tree for risk tier, and point-by-point tallying against the ten published framework points (see the framework). There is no machine-learning model involved, and your answers are never used to train one. The same answers always produce the same result, which is what makes it auditable.Weighting, in outline. Each of the ten framework points contributes to the overall readiness score; points are evidenced individually (one combined proof document each) and assessed against the same published standard regardless of which tier you're applying for. We don't publish the exact scoring thresholds and point-weightings themselves — the same reason exam boards don't publish a marking key in advance — but the assessment isn't a black box either: what's measured (the ten points), how evidence is submitted (one proof document per point), and how a result is reached (a human validates against the published framework, not a hidden model) are all public.A self-assessment result is always provisional. It never becomes a Charter Mark by itself.

Human oversight of the actual decision

We can state this with confidence: no certification decision is ever based solely on AI. No automated system decides a certification outcome. A Charter Mark is only ever issued after a named human-validation step: Sacred Logic reviews the submitted evidence against the framework and records one of four outcomes — certified, certified with conditions, deferred, or declined. See Certification Rules § 3–5 for the full process, and § 4 for who performs that review today.This matches our Privacy Policy's automated decision-making disclosure: no decision producing a legal or similarly significant effect is made solely by automated means.

The chat assistant — this is AI

The floating chat widget on every public page is a genuine AI system: it sends your message to Google's Gemini API (currently the free tier) along with a fixed knowledge base about Sacred Logic, and returns Gemini's reply. It cannot certify, decline, suspend, or revoke anything — it answers questions.It is instructed never to claim a specific named organisation is certified (it points you to /verify or /registry instead), never to give legal, financial, or regulatory advice, and to decline questions unrelated to Sacred Logic rather than improvise. It can be wrong, and is instructed to say so.Full detail on what data it processes and how long it's kept is in the Privacy Policy's AI assistant (chatbot) section.

Limitations

Neither system is infallible. The self-assessment score is only as good as the evidence you provide and is explicitly provisional; the chat assistant is a general-purpose language model answering from a fixed knowledge base and can misstate something. Neither is a substitute for reading the underlying policy pages or, for anything with legal or financial consequences, taking your own professional advice.