Singapore – Singapore has introduced three initiatives to strengthen artificial intelligence (AI) safety, improve testing standards and bolster cybersecurity as AI systems become more autonomous.
The Infocomm Media Development Authority (IMDA) announced the measures on 14 October, including a starter kit for testing AI agents, an accreditation programme for third-party AI testing providers and a new taskforce focused on AI-enabled cyber threats.
The initiatives aim to help organisations assess the safety and reliability of AI applications, establish stronger governance practices and improve the resilience of Singapore’s infocomm and media sector against emerging cybersecurity risks.
Starter kit provides guidance for testing AI agents
The new Starter Kit for Testing Agentic Systems for Safety and Reliability provides organisations with practical guidance for evaluating AI agents, which can reason, act on behalf of users, access sensitive information and make changes to their operating environments.
Developed using input from government agencies and more than 20 industry deployers and testers, the guide consolidates emerging practices for assessing agentic AI systems. It also draws on testing experience from local AI safety initiatives and includes case studies from organisations such as Cohere, CyberSierra, Earlybird AI, KBTG and Knovel Engineering.
The starter kit outlines a four-step approach to testing agentic systems: identifying potential risks, developing realistic tests, running tests and analysing logs, and implementing continuous monitoring and retesting.
It also addresses agent-specific considerations, including the accuracy of tool use and hand-offs between systems, the frequency of human intervention, and the analysis of logs to identify potential failures.
The guidance is designed to help organisations adapt their testing methods to different levels of access to an AI system, including situations where internal teams or external testers have limited visibility into how a system operates.
The document builds on Singapore’s existing guidance for responsible AI deployment and extends previous testing recommendations for large language model applications. It will be updated as testing practices and the technology evolve.
Accreditation programme recognises five AI testing firms
The AI Verify Foundation (AIVF) has introduced the AI Tester Accreditation Programme (AI TAP) to establish standards for third-party providers assessing AI systems across sectors, including healthcare, finance, recruitment and public services.
Companies seeking accreditation undergo technical, operational and financial assessments covering their testing capabilities, track records, service delivery and financial sustainability. The programme is intended to help businesses identify qualified testing providers and gain greater confidence in the assessment of AI safety and reliability risks.
The first five accredited firms are AIDX Tech, Knovel Engineering, PwC Singapore, Resaro and Vulcan. Accreditation is valid for one year, allowing providers to undergo regular reviews as AI technologies and testing requirements change.
The programme is intended to strengthen independent AI assurance by establishing a pool of testing providers capable of evaluating systems against relevant safety and reliability requirements.
New taskforce to address AI-enabled cyber threats
IMDA has also formed the Frontier AI Resilience and Security Taskforce (FAIRS) to strengthen the infocomm and media sector’s defences against cyber threats enabled or accelerated by AI.
The taskforce brings together more than 20 leaders and experts from telecommunications, media, cybersecurity and AI. It will focus on improving the sector’s ability to respond to AI-accelerated attacks, strengthen the resilience of networks and systems, and share intelligence on emerging threats.
Its work will cover three priorities: improving network visibility and incident response capabilities; examining ways to strengthen and segment networks to limit vulnerabilities and disruption; and pooling industry expertise to identify new attack techniques and develop practical guidance.
The initiatives form part of Singapore’s broader approach to supporting AI adoption while strengthening governance, accountability and cybersecurity. By combining practical testing guidance, independent assessment and sector-wide collaboration, the measures are intended to help organisations manage risks as AI becomes more deeply integrated into business operations and essential services.
