The Model AI Governance Framework: Singapore's rulebook, from 2019 to AI agents
Singapore's framework family from 2019 to the agentic rules, mapped.
Published 6 October 2026
Long-form examination of the law that governs AI agents, and what it asks of the teams that build and ship them. Each piece states the date it is current as of, cites primary text, and is revised when the law moves rather than left to age quietly.
Singapore's framework family from 2019 to the agentic rules, mapped.
The US position on agent liability, and the records that decide it.
NYC Local Law 144 after the Comptroller's December 2025 audit. Who can perform the bias audit, what must be published, and why an annual audit fits an AI agent badly.
When an AI agent causes an incident, the response fails on reconstruction, not detection. What your logs must answer, grounded in the EU AI Act, DPDP, DIFC and NIST.
What an AI agent audit trail must contain, derived field by field from the EU AI Act, DPDP Rules 2025, DIFC Regulation 10, NIST, ISO 42001 and IMDA.
DIFC Regulation 10 requires deployers of autonomous systems to produce evidence on request, not just records. What Regulation 10.2.2(c) to (g) actually demand of an AI agent.
IMDA's Model AI Governance Framework for Agentic AI, now at Version 1.5, is the world's first governance framework for AI agents. What its four dimensions ask of a deployment.
The Digital Omnibus deferred the EU AI Act's high-risk regime but not its transparency duties. Which obligations bind AI agents now, and which arrive in 2027 and 2028.
EU AI Act Article 14 human oversight translated into agent architecture: the five capabilities of Article 14(4), automation bias, and what a safe halt means mid-task.
Read against a deployed AI agent, the DPDP Rules 2025 are conduct regulation. What Section 8, Rule 6 and Rule 8 require, and why 13 May 2027 is the date that matters.