Singapore AI Governance Framework AGT-2.2.1: Allocate accountability across the agent value chain, internally and in contracts

Agentic framework 2.2.1: deploying organisations and their people remain accountable for agents' actions; map roles across the value chain (model developers, platform providers, system providers or app developers, tooling providers such as MCP servers and APIs, deployers, end users); internally assign responsibilities to decision makers (use cases, data limits, governance approach), product teams (design, testing, monitoring, user education), cybersecurity teams (guardrails, red teaming, threat modelling) and users (responsible use, training, reporting issues), and build capability for adaptive governance; externally set obligations on security, performance and data protection in contracts and address third-party opacity by requiring disclosures and evaluating authentication and logging features.