Recommendation on the Ethics of Artificial Intelligence

Claims that cite this source · 7

1. Interpretation· confidence medium· passage checked

   Accountability overlaps with answerability, responsibility, liability and remedy, but those concepts are not interchangeable.

   Used in: Accountability, Contestable decision

2. Source assertion· confidence high· passage review pending

   Human rights and dignity can function as the normative foundation against which AI governance choices are evaluated.

   Locator: UNESCO values and principles; EU AI Act Article 1

   Used in: Affected people, Impact assessment

3. Interpretation· confidence medium· passage review pending

   Human oversight is a context-dependent safeguard; merely placing a person in a workflow does not establish that intervention is informed or effective.

   Used in: Meaningful human oversight

4. Synthesis· confidence medium· passage review pending

   People affected by AI systems are governance participants and rights-holders, not only downstream recipients of technical risk controls.

   Used in: Affected people, Civil society and independent researchers, Contestable decision

5. Synthesis· confidence high· passage checked

   UNESCO connects accountability to rights, oversight and access to remedy, not only to internal documentation.

   Used in: Complaint and appeal routes, Accountability

6. Interpretation· confidence medium· passage checked

   A principle states a value; in this framework, a safeguard is a specified mechanism with an actor, a trigger, a record and a consequence. Specifying the mechanism lets us inspect how the value reaches a decision.

   Used in: Internal safety and governance team, Pre-deployment evaluation and red teaming, Safety

7. Interpretation· confidence medium· passage checked

   Most definitions aim at risk, compliance, or responsible and ethical use; several of them, by their wording, operate system by system. AIGovSpec aims at harmful patterns. The closest counterparts are UNESCO’s call for anticipation and monitoring of impact and NIST’s view that governance is continual over a system’s lifespan, but neither makes patterns, as distinct from individual risks, the target.