Skip to main content
AI in Asia
Policy

Singapore Files First GenAI Testing Standard

Singapore-led ISO/IEC 42119-8 is the first international GenAI testing standard - Asia gets a permanent seat at the rule-writing table.

· Updated Apr 26, 2026 7 min read
Singapore Files First GenAI Testing Standard

Singapore Just Filed The First Global GenAI Testing Standard, And Asia Now Owns A Piece Of The Rulebook

Singapore's Infocomm Media Development Authority has formally introduced ISO/IEC 42119-8 as the first international standard targeting generative AI testing methodology, including red teaming and benchmarking. The submission, made public earlier in April, lands at the same moment Prime Minister Lawrence Wong is standing up a National AI Council and committing the city-state to four sectoral AI Missions. Together those moves mean Asia is no longer just adopting AI rules written elsewhere. It is writing them.

What The Standard Actually Specifies

ISO/IEC 42119-8 is the eighth part of the broader 42119 series and covers methodology for testing generative AI systems. The Singapore-led draft sets out a structured approach to benchmarking outputs, defining red-team scenarios, and reporting evaluation results in a comparable way across vendors. The full text has not yet been ratified, but IMDA's announcement confirms Singapore is the convenor of the working group and that the document is now in the formal balloting stage at the International Organization for Standardization.

The standard matters because GenAI procurement, especially in regulated Asian sectors like banking, healthcare, and government, has been operating without an internationally agreed evaluation language. A bank in Jakarta currently has no neutral basis on which to compare claims made by Anthropic, OpenAI, Alibaba, and DeepSeek about safety and refusal behaviour. ISO/IEC 42119-8 attempts to fix that.

Singapore has worked very deliberately to position itself as the place where AI safety standards are written, not just enforced. This is the most concrete output of that strategy yet.

Aaron Maniam, Vice Dean for Public Sector Leadership, Lee Kuan Yew School of Public Policy

How It Fits The Wider National AI Council Reset

Singapore's policy posture shifted in February 2026 when Prime Minister Wong announced the National AI Council, which he chairs personally. The council has commissioned four sector-specific AI Missions covering advanced manufacturing, connectivity and logistics, finance, and healthcare, and the Champions of AI programme adds enterprise transformation and workforce training co-funding for participating firms.

The combined package gives Singapore three layers of policy posture. Domestically, the AI Missions are designed to spur adoption inside priority sectors. Regionally, the Smart Nation strategy provides the diplomatic framing for ASEAN coordination. Internationally, the ISO submission gives Singapore a seat at the standards-setting table that has, until now, been dominated by US and European voices.

By The Numbers

42119

8 part of the 42119 series

8 part of the 42119 series, with Singapore-led 42119-8 specifically covering generative AI testing methodology

$5.5 billion

$5.5 billion is the Microsoft committed Singapore AI

$5.5 billion is the Microsoft committed Singapore AI infrastructure spend through 2029, the financial backdrop to the policy push

10

10 major Asian economies that now operate dedicated

10 major Asian economies that now operate dedicated AI legislation or comprehensive national strategies, per the Digital in Asia 2026 policy tracker

64%

64% of Singaporean universities now run AI virtual

64% of Singaporean universities now run AI virtual tutors, the highest rate in the eight-system Times Higher Education Asia survey

Why Asia's Policy Posture Is Quietly Winning

The Brussels-Washington-Beijing triangle has dominated Western coverage of AI governance for two years. The reality across Asia is more interesting. Japan has taken an explicitly AI-friendly stance designed to attract investment and re-energise its industrial base. South Korea passed an AI Basic Act earlier this year and has already started refining it. China has tightened cybersecurity law and is aggressively promoting Huawei-anchored sovereign stacks. Vietnam is implementing a risk-based framework loosely modelled on the EU AI Act. Singapore has chosen a third option: convene the standards-setting work itself.

That third path matters because standards travel further than statutes. A statute in Singapore binds firms operating in Singapore. An ISO standard published from Singapore can be cited in procurement contracts in Indonesia, Vietnam, Thailand, the Philippines, and beyond, without any further legislative action. That gives Singapore disproportionate influence over how AI is bought and tested across Southeast Asia.

Standards are how you export regulatory philosophy without exporting regulation. Singapore understands this better than most jurisdictions.

Lim Sun Sun, Vice President of Partnerships and Engagement, Singapore Management University

What Vendors Will Have To Change

Vendors selling into Singapore-regulated industries will be expected, almost immediately, to provide ISO/IEC 42119-8-aligned evaluation reports. That formalises what some buyers, especially Singapore's three local banks, have already been requesting on an ad hoc basis. The ripple effect is what matters. Once DBS, OCBC, and UOB request the standard, regional banks, including Maybank, Bangkok Bank, and Bank Mandiri, will mirror it inside two quarters because their procurement teams already track Singapore as a leading indicator.

For the vendors themselves the work is real but manageable. OpenAI and Anthropic already publish detailed safety reports and will need to map existing evaluations to the standard's structure. Alibaba Qwen and DeepSeek will face a steeper curve because their published evaluations are less consistent and less independently audited. The standard could therefore widen, not narrow, the gap between US-aligned and China-aligned vendors in Asian regulated procurement, even though the standard itself is jurisdiction-neutral.

Asia AI Policy PosturePrimary Tool2026 DirectionCross-Border Reach
SingaporeISO standards + AI CouncilConvening international rulesHigh
JapanPro-investment policy stanceMost AI-friendly nationMedium
South KoreaAI Basic Act enforcementRefining law, three months inLow
ChinaAmended cybersecurity lawTightening, sovereign stackMedium via Belt and Road
VietnamRisk-based frameworkEU-style implementationLow

For background, our coverage of the Korean AI Basic Act enforcement and the Taiwan AI Basic Act enforcement update shows how the rest of the region is now in active rule-making mode.

The AIinASIA View: The ISO/IEC 42119-8 submission is the policy story Asia analysts should be tracking, not the headline-friendly statutes. By writing the testing methodology rather than just adopting somebody else's, Singapore turns itself into a permanent reference point for how AI is bought across the region. The shift is structural and durable, and it forces every vendor selling into Asia to engage with a Singapore-anchored evaluation framework rather than a Brussels or Washington one. Expect the standard to be informally adopted across ASEAN procurement processes well before formal ratification at ISO is complete.

Frequently Asked Questions

When will ISO/IEC 42119-8 be ratified?

The standard is currently in the ISO balloting stage and ratification timing depends on member-body comments. Industry watchers expect a final published standard within 12 to 18 months, although Singapore-anchored procurement may begin referring to the draft text well before that.

Does this replace the EU AI Act for Asia?

No. ISO/IEC 42119-8 is a testing methodology standard, not a comprehensive regulatory framework. It can be used inside any regulatory regime, including Vietnam's EU-style framework or Singapore's AI Verify approach.

How does the standard affect Chinese AI vendors selling into Singapore?

Chinese vendors will need to provide evaluation reports aligned with the standard. The work is not impossible, but it requires more consistency and third-party auditability than some Chinese vendors currently publish. Expect a temporary widening of the procurement advantage for US-aligned vendors in regulated Singapore deployments.

Is this related to AI Verify?

Yes, indirectly. Singapore's AI Verify framework gave IMDA the institutional credibility and technical staff to convene international standards work. ISO/IEC 42119-8 builds on the ideas formalised in AI Verify and brings them into the global standards system.

Will other Asian governments push their own ISO submissions?

Probably. Japan, South Korea, and India have all signalled interest in standards work, and the Singapore precedent gives them a template. Watch the next ISO/IEC AI subcommittee meeting in Q3 for a wave of follow-on proposals.