UNESCO Recommendation on the Ethics of AI P2: para 27 Safety and security: avoid and eliminate safety risks and vulnerabilities to attack across the life cycle

Accidental harm (safety) and exposure to attack (security) are to be designed out, and dealt with where they arise, at every stage of AI; good-quality data made available through sustainable, privacy-respecting access arrangements supports better training and validation of models.

Maintained by Gerard BlokdykVerified against the published standard 31 May 2026Control text last updated 21 May 2026

Other controls in III.2 Principles (paras 25 to 47) – UNESCO Recommendation on the Ethics of AI

P1 paras 25, 26 Proportionality and do no harm: justify the use and method of AI, assess risk, keep final human determination, no social scoring or mass surveillance
P10 paras 46, 47 Multi-stakeholder and adaptive governance: respect data sovereignty, involve stakeholders across the life cycle and respect Indigenous data self-governance
P3 paras 28 to 30 Fairness and non-discrimination: share benefits inclusively, minimise bias, provide effective remedy and close digital divides
P4 para 31 Sustainability: continuously assess the human, social, cultural, economic and environmental impact of AI against sustainability goals
P5 paras 32 to 34 Privacy and data protection: handle data lawfully across its life, with multi-stakeholder governance and privacy impact assessment by design
P6 paras 35, 36 Human oversight and determination: responsibility always attributable to people or legal entities; life and death decisions not ceded to AI
P7 paras 37 to 41 Transparency and explainability: tell people when AI informs decisions, give reasons and a human review route, and explain lasting impacts
P8 paras 42, 43 Responsibility and accountability: attribute responsibility to AI actors by role, with oversight, audit, due diligence and whistle-blower protection

Query this from an agent

The graph holds this control, the 0 it maps to, and the evidence behind each claim, over MCP and REST.

MCP endpoint All UNESCO Recommendation on the Ethics of AI controls