IMDA and Microsoft join forces on AI safety, agentic AI research

The Infocomm Media Development Authority (IMDA) of Singapore and Microsoft Corporation have signed a memorandum of understanding (MOU) to deepen collaboration on artificial intelligence safety and security. The partnership focuses on three key areas: technical and research collaborations on AI safety and security, facilitating information sharing between the organizations, and developing a policy framework for trusted access to frontier AI models. A notable component is joint research into agentic AI — autonomous systems that can take actions — and the development of evaluation methods, tools, and benchmarks for AI models, including multilingual AI safety. The collaboration also seeks to strengthen societal resilience against broader trust and safety challenges posed by AI systems.

Under the MOU, IMDA, the Singapore AI Safety Institute, and Microsoft will explore how governments and infrastructure operators can responsibly structure access to frontier AI models for safety testing. This work includes co-developing a white paper examining demand-side needs (e.g., government agencies) and supply-side policy considerations for model providers. Kiren Kumar, Deputy Chief Executive of IMDA, emphasized that this goes beyond policy frameworks to jointly building benchmarks and tools, advancing the science of evaluation. Natasha Crampton, Microsoft's Chief Responsible AI Officer, noted that Singapore is shaping global conversations on safe AI deployment, and this partnership combines government insight with Microsoft's technical and operational experience to address emerging risks and build trust.