UNESCO
Normative ethics grounded in human rights, dignity, transparency, fairness, and human oversight.
- Audience
- Member states, policymakers, organizations.
- Unit of analysis
- Principles and values.
- Lifecycle coverage
- Whole-of-system, normative.
- Outputs
- Policy commitments; values statements.
- Strengths
- Globally negotiated; rights-based; broad legitimacy.
- Cautions
- High-level; not directly operational; not a control framework.
- Jurisdictional scope
- Global · 193 member states
- Evidentiary weight
- Normative; influences national legislation.
- Cost to adopt
- Low for declaration; high for full Readiness Assessment.
- Certification path
- No certification; UNESCO Readiness Assessment Methodology (RAM).
Adopted unanimously by 193 member states at the 41st session of the General Conference, November 2021. First global standard-setting instrument on AI ethics.
Recommendation on the Ethics of Artificial Intelligence
Indexed at the structural level. Excerpts are quoted under fair-use; full text is linked, not rehosted.
Principles10
- Human oversight & determinationframingdeploymentmonitoring
Human oversight and determination
“Member States should ensure that it is always possible to attribute ethical and legal responsibility for any stage of the life cycle of AI systems to physical persons or to existing legal entities.”
- Fairness & non-discriminationdatamodeldeploymentmonitoring
Fairness and non-discrimination
“AI actors should promote social justice, fairness and non-discrimination of any kind in compliance with international law.”
- Transparency & explainabilitymodeldeploymentmonitoring
Transparency and explainability
“The transparency and explainability of AI systems are often essential preconditions to ensure the respect, protection and promotion of human rights, fundamental freedoms and ethical principles.”
- Privacy & data protectiondatamodeldeploymentretired
Right to privacy and data protection
“Privacy must be respected, protected and promoted throughout the AI system life cycle.”
- Safety & securitymodeldeploymentmonitoring
Safety and security
“Unwanted harms (safety risks) and vulnerabilities to attack (security risks) should be avoided, addressed and eliminated throughout the life cycle of AI systems.”
- Environment & ecosystemsframingdatamodeldeployment
Environment and ecosystem flourishing
“AI technologies should contribute to environmental and ecosystem flourishing.”
- Proportionality & do-no-harmframing
Proportionality and do no harm
“AI methods should not exceed what is necessary to achieve a legitimate aim, and should be appropriate to the context.”
- Responsibility & accountabilityframingdeploymentmonitoringretired
Responsibility and accountability
“AI actors and Member States should respect, protect and promote human rights and fundamental freedoms, and should foster the protection of the environment and ecosystems.”
- Awareness & literacyframingdeployment
Awareness and literacy
“Public awareness and understanding of AI technologies and the value of data should be promoted through open and accessible education, civic engagement, digital skills and AI ethics training.”
- Multi-stakeholder governanceframingdeploymentmonitoring
Multi-stakeholder and adaptive governance
“Participation of different stakeholders throughout the AI life cycle is necessary for inclusive approaches to AI governance.”
Policy areas05
- Policy Area 1framing
Ethical impact assessment
“Member States should introduce frameworks for impact assessments to identify and address impacts on human rights, the rule of law, democracy, and ethical considerations.”
- Policy Area 2framingdeploymentmonitoring
Ethical governance and stewardship
“Adopt regulatory frameworks that translate ethical values into actionable policy, with mechanisms for redress.”
- Policy Area 3data
Data policy
“Develop data governance strategies that ensure quality, security, privacy, and openness consistent with international human rights law.”
- Policy Area 8framingmodeldeploymentretired
Environment policy
“Assess environmental impacts of AI systems including the carbon footprint and consider full life cycle effects.”
- Policy Area 6datamodeldeployment
Gender
“Ensure that gender equality is mainstreamed in policies and that AI does not reinforce existing gender biases.”