UNESCO Recommendation on the Ethics of 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

Transparency and explainability underpin rights, liability and the ability to challenge decisions, at a level suited to context and impact and balanced against privacy, safety and security. Anyone subject to a decision that AI made or informed should be told so, be able to ask for an explanation, access reasons for decisions affecting their rights, and make submissions to a designated person able to review and correct the decision; users should be told promptly when a product or service uses AI. Where serious rights impacts are threatened, sharing code or datasets may be needed. Algorithms should be explainable, and decisions with lasting, not easily reversible or not low-risk effects should come with a meaningful explanation.