IMDA released Version 1.5 of the Model AI Governance Framework for Agentic AI on May 20 2026 - 51 pages incorporating feedback from over 60 organisations including GovTech DBS AWS Google OCBC and Tencent. The AI Asia Pacific Institute assessed the framework as sitting at an important midpoint between high-level principles and formal system-theoretic governance. Version 1.5 introduces dedicated discussion of systemic risks including agent sprawl miscoordination conflict collusion and emergent behaviours and acknowledges cascading effects and unpredictable outcomes as documented risks. Organisations remain legally accountable for agent behaviours regardless of voluntary MGF compliance. No component of Version 1.5 specifies a pre-execution constitutional gate that physically prevents cascade failure propagation across Singapore's nationally dense infrastructure before consequences become irreversible. This paper documents the structural gap between IMDA's world-leading governance framework and the system-theoretic constitutional enforcement architecture required to complete it.
Read paper: https://doi.org/10.5281/zenodo.20683565
For full paper and sovereign briefings: protocol@fijishi.com