The Human Values Framework
© Sacred Logic™ 2026
Every Sacred Logic™ Charter Mark is assessed against these ten points — and human-reviewed before it's awarded. Always. Points 1–7 align with the EU's own Trustworthy AI requirements; points 8–10 go further, covering authorship, cultural diversity, and whole-of-life stewardship that conventional frameworks leave out.
1. Human Agency and Oversight
Meaningful human decision-making authority is preserved. AI decisions with material or legal effects on people have a competent human review layer — with real checkpoints, escalation, and override procedures.
2. Technical Robustness and Safety
The system is resilient against error, failure, and adversarial manipulation, with a deterministic fallback when things go wrong — and a tested incident-response and disclosure plan.
3. Privacy and Data Governance
A verified lawful basis for processing, data minimisation, security by design, and working pathways for people to exercise their data rights.
4. Transparency and Explicability
People are told when they're interacting with or being assessed by AI. Synthetic content is disclosed and labelled, and the system's logic is documented in proportion to its risk.
5. Accountability and Governance
Explicit accountability across development and deployment, an accessible complaints-and-remediation pathway, and an independent appeals process.
6. Fairness, Non-Discrimination and Accessibility
Systematic insulation against bias across protected characteristics, evidenced by dated, repeatable bias audits — plus universal design for accessibility (WCAG 2.2).
7. Societal and Environmental Well-Being
The system's societal impact and environmental footprint — including compute intensity — are assessed and disclosed, with genuine stakeholder consultation.
8. Creative and Intellectual Authorship Integrity
AI involvement in creative and intellectual outputs is disclosed. Attribution and moral rights are preserved, and AI output is never misrepresented as unaided human work.
9. Cultural and Epistemic Diversity
The risk of epistemic homogenisation is assessed. Linguistic and cultural pluralism is reflected in data and outputs, with multi-stakeholder input on cultural impact.
10. Proportionality and Lifecycle Stewardship
Controls are proportionate to the system's actual risk tier, with periodic re-assessment, a change log, and a defined decommissioning procedure — stewardship for the system's whole life, not just its launch.
Already comply with the EU AI Act? Use the Compliance Fast Track
You provide one combined proof document per point, not a long checklist. And if you already hold EU AI Act conformity — a CE mark, an ISO/IEC 42001 certificate, or a signed self-declaration — the Compliance Fast Track has you complete only points 8–10. Points 1–7 are taken as met by your existing conformity, which we check documentarily rather than re-audit. Your public register entry shows the basis openly.
Free proof-document template pack
A ready structure for the one combined proof document each point needs — what to include, plus an evidence-record table to complete. Download the template pack (PDF) ↓
Want to see where your AI system stands against these points? Run the free 5-minute self-assessment — no signup required.
Every Charter Mark human-reviewed. Always. · Built on a 10-point Human Values Framework · View the public registry
The Charter Mark is a voluntary Article 95 code-of-conduct certification, not a statutory EU AI Act conformity assessment, and is separate from any mandatory conformity assessment your system may require.