Trust is the infrastructure of AI adoption. This session focuses on what it takes in 2026 to govern AI for real world benefit, especially in the Global Majority. We move from principles to practice through sandboxes with clear entry and exit criteria, independent evaluation and red teaming, and reusable assurance artifacts such as model cards, audit logs, and incident reporting. We ask where interoperability and collaboration accelerate impact, and where sovereignty and geopolitics create friction.