무슨 일이 있었나
On 14 October 2026 at the AI Assurance Exchange (Agentic Edition, SICW) in Singapore, IMDA and the AI Verify Foundation launched: (1) the 'Starter Kit for Testing Agentic Systems for Safety and Reliability' — a first government-issued methodology to identify agentic AI risk, design tests, analyse logs, monitor ongoing behaviour and validate safeguards across four phases (risk identification, test design, results analysis, ongoing monitoring); (2) the AI Tester Accreditation Programme (AI TAP), announcing the first five accredited third-party AI testing providers (AIDX Tech, Knovel Engineering, PwC Singapore, Resaro, Vulcan); and (3) a Frontier AI Resilience and Security Taskforce to strengthen the infocomm/media sector's resilience against AI-enabled threats. Developed with 20+ industry deployers/testers (Google, AWS, Microsoft, Singtel) and Singapore's AI Safety Institute.
왜 중요한가
This is a coordinated national-government response to the agentic-AI safety gap, operationalising testing of autonomous agents (prompt injection surface, authorization-scope ambiguity, non-binary safety properties, behavioural drift). It extends Singapore's Model AI Governance Framework for Agentic AI and sets a de facto baseline for agentic AI testing and third-party assurance, with a credentialing programme that will shape the global AI safety-testing market.
필요한 조치
Organisations deploying agentic AI systems should use the Starter Kit to structure pre-deployment and ongoing agent testing; AI safety-testing/assurance firms should evaluate AI TAP accreditation requirements.