OpenAI opens Singapore lab; IMDA revises agentic AI rules

OpenAI announced it will open an Applied AI Lab in Singapore under a partnership with the Ministry of Digital Development and Information, backed by a commitment of more than S$300 million. The initiative, revealed at the ATx Summit, will create more than 200 Singapore-based technical roles over the next few years and make the city-state a hub for forward-deployed engineers who work with organisations on AI deployment.

The lab’s work will align with Singapore’s AI Mission priorities, including public service, finance and digital infrastructure. OpenAI will collaborate with government agencies and local partners on education and workforce programmes linked to the Ministry of Education and GovTech.

OpenAI plans a Singapore chapter of the OpenAI Academy, participation in the National AI Impact Programme, and Codex for Teachers hackathons. The company expects to run accelerator-style workshops for AI-native startups and training for micro-entrepreneurs and small businesses on using AI in operations and customer service.

Chng Kai Fong, Permanent Secretary for Digital Development and Information, described Singapore’s response to AI as “growing new sectors, anchoring global frontier companies, and equipping workers with relevant skills.”

Separately, the Infocomm Media Development Authority released an updated governance framework for agentic AI. The revision builds on Singapore’s 2020 Model AI Governance Framework and the agentic guidance first published in January 2026. IMDA revised the framework after consulting more than 60 organisations and expanded guidance on risks tied to multi-agent systems, third-party agents, automation bias and human accountability.

The updated framework includes more than ten case studies from Singaporean and international organisations. One case study from Dayos describes an AI-powered ticketing agent for internal IT requests. Dayos uses tiered risk levels: low-risk reversible tasks such as password resets are automated and audited every two weeks; moderate-risk actions require human approval; higher-risk actions with limited reversibility are excluded from the agent’s authority.

A Tencent case study covers CodeBuddy, an agentic coding system that can plan, write and deploy code, access filesystems, run terminal commands and call external APIs. CodeBuddy enforces preset defaults and configurable permissions and requires human approval for editing files, running shell commands, making network requests or invoking external tools. The system offers plain-language explanations of complex commands and flags suspicious commands for review.

GovTech Singapore documented a phased rollout of agentic coding assistants restricted to GovTech employees, barred external tools initially and limited use to low-risk systems. The agency set up central logging, a framework for connecting approved external tools and tests to assess resilience to potential attacks.

OpenAI’s funding commitment exceeds S$300 million and the lab will add over 200 local technical jobs; IMDA’s revised agentic framework adds case studies and new guidance for organisations deploying autonomous AI agents.