Singapore has unveiled three initiatives to strengthen the safety, assurance and cyber resilience of artificial intelligence, with a particular focus on the fast-emerging risks posed by agentic AI systems that can reason, act and make decisions on behalf of users.
Announced at the AI Assurance Exchange: Agentic Edition on 14 October, the measures include a Starter Kit for Testing Agentic Systems for Safety and Reliability, the launch of the AI Tester Accreditation Programme (AI TAP), and the formation of the Frontier AI Resilience and Security Taskforce (FAIRS). The event, organised by the AI Verify Foundation and SGTech, brought together more than 300 AI deployers and testers.
Assurance for agentic AI
Unlike traditional or generative AI, agentic AI can access sensitive data, use external tools, alter systems and execute tasks autonomously. Singapore’s new Starter Kit provides a four-step approach: identifying risks such as erroneous or unauthorised actions; designing realistic tests; running tests and analysing logs; and implementing continuous monitoring and re-testing.
The guidance was developed through consultations with government agencies and more than 20 industry deployers and testers, including Google, AWS, Microsoft, Singtel, Resaro and Workday. It also incorporates case studies from organisations including Cohere, CyberSierra, Earlybird AI, Kasikorn Business-Technology Group (KBTG) and Knovel Engineering.
In his opening address, Infocomm Media Development Authority (IMDA) chief executive Ng Cher Pong said the challenge is to ensure AI’s growing autonomy is matched by stronger controls. “When we give AI more agency, we need greater assurance over how that agency is exercised,” he said.
Ng added that testing must assess not only whether an agent reaches the right outcome, but also which tools it used, what information it accessed, whether it respected permissions, and whether safeguards worked as intended.
Accredited testing ecosystem
The AI Verify Foundation also launched AI TAP, described as the first accreditation programme of its kind for third-party AI testing providers. Firms seeking accreditation undergo technical, operational and financial evaluations covering testing competency, track record, delivery capability and financial sustainability.
The first five accredited firms are AIDX Tech, Knovel Engineering, PwC Singapore, Resaro and Vulcan. Accreditation lasts one year, allowing standards to be reviewed as AI technology evolves.
The programme is intended to support organisations across sectors including healthcare, finance, hiring and public services. “As AI moves from informing decisions to executing them, organisations need independent evidence that these systems can be trusted in practice,” said April Chin, co-chief executive of Resaro.
Regional cyber resilience
The third initiative, FAIRS, brings together more than 20 leaders and experts from telecommunications, media, cybersecurity and AI to strengthen Singapore’s Infocomm and Media sector against AI-enabled cyber threats. The taskforce will publish practical guidelines and best practices by the first half of 2027.
Its priorities include improving network visibility and response capabilities against AI-accelerated threats; strengthening and segmenting critical systems as AI becomes better at identifying vulnerabilities; and pooling sector-wide intelligence to detect emerging attack techniques.
For Asian enterprises, the announcements signal a shift from AI governance principles towards operational assurance: testing systems across their full decision-making trajectory, engaging accredited independent testers, and preparing critical infrastructure for AI-driven cyber risks.
Singapore’s approach may offer a practical reference point for organisations across Southeast Asia as agentic AI moves from pilot projects into customer service, internal productivity and complex business processes.
