Human oversight as a procedural safeguard under the Council of Europe AI Framework Convention

The contribution explores the possibility of identifying an obligation of human oversight of AI systems within the Council of Europe Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law. Although the final text of the Convention does not explicitly impose such an obligation, the need (or at least the desirability) to ensure human oversight played a central role throughout the entire negotiation process and is still mentioned in the Explanatory Report. After retracing the main stages of how the issue was addressed during the negotiations, the contribution focuses on the interpretative value of the reference to human oversight contained in the Explanatory Report in the light of Articles 31 and 32 of the Vienna Convention on the Law of Treaties, concluding that, while it does not qualify as an "authentic" (and thus legally binding) interpretation, the reference to human oversight in the Report provides valuable interpretative guidance. It will also be argued that an obligation of human oversight may nevertheless be derived from the broader human rights framework that the Convention is intended to implement in the AI domain, having particular regard to States’ positive obligation to protect individuals against human rights violations. Finally, the article identifies a number of practical conditions for human oversight to operate effectively as a safeguard mechanism, with particular attention to system design, operator training, and the organisational context.