From Allocation to Answerability: A Feedback Submission on the Model AI Governance Framework for Agentic AI (Version 1.5)

This paper is a feedback submission to the Infocomm Media Development Authority of Singapore, responding to its call for comment on the Model AI Governance Framework for Agentic AI (version 1.5). The framework is operationally strong, but at three points it secures a proxy for the property it sets out to protect — ‘meaningful accountability’ — so that a fully compliant deployment can still leave no human able to answer for what an agent did. The submission sets out three refinements, summarised below and developed against specific dimensions of the framework. Each is anchored, where possible, to the framework’s own text, and each is framed as an addition that builds on the framework’s existing commitments rather than a replacement for them.