OpenAI Opens AI Lab in Singapore

OpenAI Establishes Presence in Singapore: An AI Lab to Revolutionize the Tech Sector

OpenAI has announced the opening of its first applied artificial intelligence laboratory outside the United States, and it is happening in Singapore. This laboratory is part of a strategic partnership with the Ministry of Digital Development and Information of Singapore.

This initiative, dubbed "OpenAI for Singapore," was unveiled at the ATx summit, marking a financial commitment of over 300 million Singapore dollars. This ambitious project aims to create more than 200 technical positions in the city-state over the coming years. OpenAI specified that Singapore will become one of its main global hubs for on-ground engineers, who will collaborate with various organizations for AI deployment. The lab's activities will align with the priorities of Singapore's AI Mission, which includes sectors such as public service, finance, and digital infrastructure.

Focus on Deployment and Talent

OpenAI will work closely with government agencies and local partners to develop educational and vocational training programs, in partnership with the Ministry of Education and GovTech. The company also plans to support teachers through a Singapore chapter of the OpenAI Academy, participate in the National AI Impact Program, and organize "Codex for Teachers" hackathons.

The partnership also includes initiatives to collaborate with local partners on acceleration programs for AI-focused startups. These programs will take the form of workshops aimed at micro-entrepreneurs and small businesses, showing them how to integrate AI into their operations and enhance their customer service.

Chng Kai Fong, Permanent Secretary for Digital Development and Information, emphasized that Singapore's response to the rise of AI includes developing new sectors, attracting leading international companies, and equipping workers with skills suited to new technologies.

Singapore Updates the Governance Framework for Agentic AI

In parallel, Singapore has updated its governance framework for agentic AI, a framework initially launched by the Infocomm Media Development Authority at the World Economic Forum in January 2026. This revised framework builds on the AI governance model introduced in 2020 and provides organizations with guidelines for the responsible deployment of AI agents, including measures to mitigate risks associated with agentic AI.

The IMDA revised this framework after gathering feedback and case studies from the industry, with contributions from over 60 organizations, including AWS, DBS, Google, and Salesforce.

The new guidelines address risks related to multi-agent systems, third-party agents, automation bias, and human accountability. The framework now includes more than ten case studies illustrating how organizations have applied its recommendations.

Case Studies Illustrate Governance Controls

Among the case studies, that of Dayos, an AI automation company based in Singapore with operations in the United States, is particularly noteworthy. Dayos developed an AI-powered ticketing agent to manage internal IT requests. This agent is capable of automatically resolving certain requests and forwarding them to a human if necessary.

Dayos implemented hierarchical risk levels to determine the actions the agent could undertake. Low-risk and reversible actions, such as password resets, could be automated and audited every two weeks, while moderate-risk actions required human approval before execution. High-risk actions, such as permission changes with limited reversibility, were excluded from the agent's authority.

Tencent also contributed a case study on CodeBuddy, an agentic coding system developed by Tencent Cloud. CodeBuddy can plan, write, and deploy code via natural language instructions and can access file systems, terminal commands, external APIs, and MCP tools.

CodeBuddy uses predefined default settings and configurable permissions. Human approval is required for actions such as editing files, executing shell commands, network queries, or using external tools. The system explains complex commands in simple language before users approve them. Suspicious commands always require human approval, even if similar commands had been pre-approved.

The case study from GovTech Singapore covers the deployment of agentic coding assistants within the government. The first phase was limited to GovTech employees, did not allow external tools, and was restricted to low-risk systems. GovTech developed centralized logging and a framework to connect approved external tools. The agency also tested the system against potential attacks.