This report draws on workshops, stakeholder sessions, and in-country engagements with eleven countries and territories conducted through UNDP’s AI Trust & Safety Re-imagination Programme. It does not rank countries or propose a universal regulatory model. It surfaces what is already happening: the decisions being made, the risks materializing, and the governance approaches that are working under real institutional constraints. 

The central finding is straightforward. Many of the most consequential AI governance decisions are not being made in policy frameworks. They are being made in procurement offices, vendor contracts, software update cycles, and the informal decisions of civil servants and institutional teams navigating AI use without clear guidance, approval processes, or accountability pathways. Getting governance right means embedding it in these operational settings, not treating it as a separate layer above them.