Human Oversight in International AI Frameworks

The Language of Nine International AI Frameworks

Report Image

Our report shows how international frameworks address human oversight of AI.

Authors

Nils Deeg

Date Published

16 Nov 2026

Related Topics:

Our text-level analysis of nine international AI frameworks shows that human oversight is found across eight frameworks, only being absent from the G7 Principles. A total of 16 provisions address human oversight, with various degrees of focus and actionability.

  • Eight provisions focus on human oversight as the core concern, while eight provisions only touch upon human oversight tangentially.

  • 15 provisions across eight frameworks provide actionable measures to pursue human oversight. Only one provision, in the Council of Europe Framework Convention, addresses human oversight in non-actionable language.

  • The ASEAN Guide and UNESCO Recommendation have the most provisions on human oversight (4 each), all of which are actionable. Further, the UNESCO Recommendation also has the most core provisions on human oversight (3).

The report outlines the prevalence, focus, and actionability of relevant provisions in each international framework, before summarising their language. It is part of our flagship report on The Fabric of AI Governance, which covers ten concerns.