Singapore Launches Three AI Safety Initiatives

Singapore's Infocomm Media Development Authority (IMDA) introduced an agentic AI testing kit, an accreditation program for AI testing firms, and a cybersecurity task force, the authority said in a press release.

The testing kit outlines four steps: identifying risks, creating realistic tests, running tests and reviewing logs, and conducting continuous monitoring and further testing. It also covers agent actions, tool use accuracy, handoffs, human involvement, and testing when organizations have limited visibility into a system. The kit builds on Singapore's Model Governance Framework for Agentic AI.

The AI Verify Foundation also launched the AI Tester Accreditation Programme for third party testing providers. AIDX Tech, Knovel Engineering, PwC Singapore, Resaro, and Vulcan received the first one year accreditations after technical, operational, and financial assessments.

IMDA also formed the Frontier AI Resilience and Security Taskforce, which includes more than 20 representatives and experts from telecommunications, media, cybersecurity, and AI. The group will publish guidance in the first half of 2027 on responding to AI enabled threats, strengthening network resilience, and sharing intelligence about new attack methods.