On 11 February 2025 Singapore’s Ministry of Digital Development and Information announced a new round of AI safety initiatives. The round concentrates on testing and evaluation: a Global AI Assurance Pilot for generative AI applications, a joint testing report with Japan, and an evaluation report from the AI Safety Red Teaming Challenge.
Why testing and evaluation
Among the many tools of AI governance, testing and evaluation sit between rules and technology. Law sets the direction of an obligation, but whether a given system meets it needs a repeatable test method to answer. Singapore’s choice is to build the method first and then have it adopted across the region.
The Global AI Assurance Pilot targets generative AI applications and promotes best practice in technical testing; the joint report with Japan seeks comparable methods across jurisdictions. The value of such work is not a single conclusion but a reusable assessment framework. See our notes on the G7 cyber expert group statement and on the NIST GenAI pilot evaluation.
What it means for companies
For AI companies operating in Singapore and Southeast Asia, standardised testing and evaluation has two effects. First, safety assessment may become a precondition for entry into certain procurement scenarios, especially in the public sector and regulated industries. Second, once a method is adopted regionally it lowers the cost of repeated self-certification — provided the company’s own test records can be understood externally.
The red-teaming evaluation report also points in a direction: adversarial testing is moving from an internal practice to a standard action open to external review. See our notes on the AI safety governance committee.
Part of a regional picture
Put back into the regional context, Singapore’s move completes a clearer line: at the G7 level, the push is toward application and diffusion; at the regional level, the work is on testing and evaluation methods. Together they form a structure in which “encouraging use” and “controlling risk” run in parallel. See our notes on the G7 AI for Prosperity statement.
For companies going overseas, the practical point is to identify, before entering a market, whether it weights application diffusion or risk control more, and to shape compliance materials accordingly.
This article is based on announcements by Singapore’s Ministry of Digital Development and Information and public information and is for reference only; it does not constitute legal advice.