Safe AI Culture: The Human-Side Governance Framework for AI Why the missing layer in responsible AI is not technical — it is human Safe AI Culture is proposed as a human-side governance framework designed to operationalise the human oversight requirements mandated by the EU AI Act, ISO/IEC 42001, and the NIST AI Risk Management Framework. The framework addresses what this paper terms Layer 4 — the people layer of AI governance: the boards, executives, and organisational conditions that determine whether technical AI governance controls function as designed or remain nominal. Drawing on 28 years of practitioner experience in cybersecurity and AI culture across multinational organisations, and on documented cases of AI governance failure in 2025–2026, the paper argues that AI adoption is following the same pattern as cybersecurity adoption two decades prior: organisations are building technical controls first and assuming the human governance layer will follow. The cybersecurity industry's experience demonstrates both the cost of that assumption and the mechanism for correcting it. The framework proposes five operational principles: Board-Level Accountability, Human-in-Command, Psychological Safety, Proportional Governance, and Stakeholder Justice. Each principle is defined operationally, evidenced by documented governance failures, and mapped to specific regulatory obligations. The paper introduces HAIRI (Human-AI Interaction Risk Index, Patent Pending EPO) as the measurement instrument that transforms each principle from aspiration into auditable evidence, and outlines a three-level certification model for organisational deployment. This working paper is the conceptual foundation of the Safe AI Culture framework. It is published as a supplement to the related book (see Related works). Author: Yolanda Mula Meneses, CEO, Security Awareness Consultants SL (SECAWACON). Board & C-Suite AI Governance Advisor. Correspondence: yolanda.mula@secawacon.com
Yolanda Mula Meneses (Wed,) studied this question.