IMDA MGF Principle on Interaction-Time Disclosure

Path Outline

Progress0%

1When a Helpful Bot Becomes an Agent 2Understanding Action Space: Tools, Not Topics 3IMDA MGF Guidance on Bounding Scope Upfront 4Case Study on Draft-Only Launch Boundaries 5Freeze a Scope Boundary You Can Defend

6Keep Scope Fixed While You Change Independence 7IMDA MGF Guidance on Meaningful Human Checkpoints 8Turn Autonomy Modes Into Checkpoints You Can Staff 9Case Study on Ticket Updates With Approval Gates 10Draft and Sign an Autonomy Profile You Can Enforce

11Why Permissions Beat Prompts for Controlling Actions 12Four Permission Classes and Their Irreversibility Risk 13IMDA MGF v1.5 on Agent Identity and Access Control 14Case Study: Scoped Write Access With Approval in Tickets 15A Permission Matrix You Can Put in Front of Security

16Turn Permissions Into Bounded Authority 17IMDA MGF on Design-Time Risk Bounding 18Build a Structural Guardrails Spec That Holds 19Defend Against Data-as-Instructions in Toolchains 20Case Study: Close Duplicates Within Guardrails

21Make Disclosure Happen Where Actions Happen

22IMDA MGF Principle on Interaction-Time Disclosure

23Trigger Points That Change User Risk 24Case Study: Add Disclosures to Ops Helper Agent

25When “The Team Owns It” Means Nobody Does 26IMDA MGF on Human Accountability for Agents 27Treating “No Owner” as a Deployment Finding 28Case Study: Owning External Follow-Ups for Ops Helper Agent

29Why a Bot Name Can’t Be Audited 30IMDA MGF on Traceable Agent Identity 31From Permission Matrix to Logs via Technical Identity 32Case Study: Registering External Actions Safely 33Handing Off an Inventory Others Can Use

34Turn Oversight Into Measurable Proof 35IMDA MGF Guidance on Monitoring and Oversight 36Case Study: Drill an Override and Rollback 37Brief Stakeholders on EU AI Act vs IMDA Guidance

Hide outline

Home
Paths
Governing Agentic AI End to End with IMDA
IMDA MGF Principle on Interaction-Time Disclosure