AI Governance / AI news for Malaysia
From the archive · Event date 28 February 2025
NAIO gathered 170 experts into seven AI working groups. The deliverables were the real test
The first meeting created a broad policy network around sovereignty, regulation, security, ethics, safety, talent and adoption. Participation was visible; the stronger test was whether the recommendations became public, traceable policy.
In brief
- NAIO said its first working-group meeting on 28 February 2025 was held online and brought together more than 170 experts from Malaysia, abroad and the Malaysian diaspora.[1][2]
- Seven groups covered AI sovereignty, regulation and policy, security, governance and ethics, safety, talent, and advisory work intended to feed Malaysia's AI Action Plan and governance models.[1][2]
- Later official updates said participation exceeded 200 and practical recommendations were produced, but the reviewed public posts did not link a complete recommendation bundle or a decision-by-decision response.[3][4]
NAIO created a broad expert network and gave it a national-policy brief
Malaysia's National AI Office moved from inviting specialists to organising them on 28 February 2025. Its first working-group meeting assembled more than 170 people spanning industry, consulting, research, professional practice, international expertise and the Malaysian diaspora. NAIO's own post said the meeting was held online.[1][2]
The structure was ambitious. Seven groups were assigned to AI sovereignty, regulation and policy, security, governance and ethics, safety, talent, and advisory work. The ministry said each group used the first meeting to set direction, objectives, deliverables and milestones, with its recommendations expected to shape the national AI Action Plan, regulatory frameworks and governance models.[1]
That announcement established a credible participation process, but it was not yet proof of policy delivery. A committee becomes publicly useful when people can inspect its outputs, understand which advice was accepted or rejected, and trace the resulting decision to a responsible agency and deadline.[1][4]
The seven groups were meant to solve different parts of the same policy problem
The sovereignty group was meant to examine Malaysia's control over AI capability, data and infrastructure. Regulation and policy focused on the national strategic direction and a workable regulatory framework. Security addressed external threats and national or digital risk, while safety focused on whether AI systems were trustworthy, reliable and aligned with the public interest.[1]
Governance and ethics dealt with how principles could be turned into operating models and guidelines. The talent group addressed the local pipeline of people able to build, deploy and supervise AI. The advisory group was expected to connect emerging trends and real implementation challenges to businesses and sector teams.[1][3]
Those boundaries matter because the labels can sound interchangeable. Security asks how systems and national assets resist threats. Safety asks how systems avoid harmful or unintended behaviour. Governance determines who decides, documents and answers for their use. Regulation determines which duties become enforceable or formally supervised.[1][3]
A later update turned the exercise into a three-month sprint with dated checkpoints
By 19 April, the Ministry of Digital said the working groups had grown to 210 contributors. It described a three-month sprint that began with direction alignment on 27 March, moved to first draft deliverables on 29 April and was due to conclude with refined outputs presented to NAIO on 29 May.[3]
The ministry said the outputs were intended to form the backbone of Malaysia's national AI roadmap. It also connected the work to practical areas: helping small and medium enterprises adopt AI, improving digital public services, strengthening responsible data governance and building local talent.[3]
NAIO later said more than 200 participants completed the March-to-May sprint and produced practical, actionable recommendations. That is evidence that the process progressed beyond its first meeting. The public post, however, announced that outcomes and next steps would be shared rather than linking the complete recommendations itself.[4]
The number of experts shows breadth, not whether every important voice shaped the result
A pool of 170 or 210 contributors can bring valuable technical and sector knowledge. It does not, by itself, show the balance between government, large technology firms, local builders, SMEs, workers, educators, civil society and communities likely to be affected by automated decisions.[1][3]
A stronger public record would include the membership or representation criteria, conflict-of-interest rules, the questions put to each group, draft and final outputs, and a short disposition table showing which proposals were accepted, changed, deferred or rejected. Sensitive security material could be summarised without exposing operational details.[1][4]
That record would also make later policy easier to evaluate. If a governance rule, safety test or talent programme appears in the national plan, readers could trace it back to the problem the group identified, the evidence it considered and the outcome the government promised to measure.[3][4]
Why Malaysia should care
A national AI plan can affect public services, business rules, cyber risk, education and jobs. Malaysians therefore need more than a large expert count: they need to see what each group recommended, who considered the advice, what changed in the final plan and which agency owns implementation.
Businesses and AI builders
The working groups were intended to influence regulation, governance, security and adoption conditions that can later affect procurement and compliance.[1][3]
Practical move: Keep data, model-risk, testing and human-oversight records now, but distinguish announced policy work from rules already in force.
Researchers, workers and civil society
The breadth of participation matters most when outside groups can see the evidence and challenge assumptions before decisions are locked in.[1][4]
Practical move: Ask for public drafts, representation criteria and a response table rather than treating the expert count as the final measure of inclusion.
Policymakers and agencies
Seven groups can surface specialised recommendations, but cross-cutting conflicts still require a named decision owner and shared implementation sequence.[3]
Practical move: Publish who resolves conflicts between safety, security, adoption, talent and sovereignty priorities, then report delivery against dated outcomes.
What Malaysians can do now
- Use the seven-group structure as a map of Malaysia's AI policy questions, not as proof that every recommendation became policy.
- Look for the recommendation documents, a government response and named delivery owners when judging whether the sprint produced accountable outcomes.
- When a later programme or rule cites NAIO's work, trace it to a dated source and confirm whether it is proposed, approved, funded or operational.
What we still do not know
The official records show a completed sprint, but not the full public decision trail.
- The full membership and representation breakdown for each working group, including conflict-of-interest arrangements.
- A complete public bundle of final recommendations and the evidence or dissent recorded behind them.
- A proposal-by-proposal government response showing which outputs entered the national plan, who owns delivery and how progress will be measured.
Sources
- 1.NAIO kicks off inaugural working group meeting, driving Malaysia's AI leadership Ministry of Digital Malaysia, 28 February 2025
- 2.NAIO convenes first-ever AI Working Groups in Malaysia National AI Office Malaysia, 28 February 2025
- 3.Malaysia moves forward with AI strategy grounded in local values and global insight Ministry of Digital Malaysia, 19 April 2025
- 4.Malaysia's AI future shaped by NAIO Working Groups National AI Office Malaysia, 10 July 2025
- 5.Momentum in motion. Strategy in progress. People at the centre. AI Malaysia / National AI Office Malaysia, 9 May 2025
