Mapping General-Purpose AI Governance in Twenty AI Middle-Power Jurisdictions

The most consequential decisions about GPAI development are being made by a small number of developers concentrated in a few countries, while the risks land globally. GPAI systems are general in their capabilities and consequential across many domains at once; several of their most significant risks (large-scale misuse, cross-border externalities, and the potential for loss of control over highly capable systems) sit upstream of any single sector or jurisdiction.

Middle powers have played significant individual roles in shaping international AI processes (the Bletchley, Seoul, and Paris AI Safety Summits; the Hiroshima Process; the Council of Europe Framework Convention; the International Network of AI Safety Institutes). Yet there is no citable structured comparative picture of what middle powers are actually doing in domestic GPAI governance that focuses on how their instruments compare, or where their approaches converge and diverge. Without that picture, it is difficult for any stakeholder – whether researchers, or decision-making actors within the jurisdictions themselves – to understand the existing landscape of GPAI governance. This research fills that descriptive gap.