AI in Malaysia

How Malaysia is governing, building and learning AI: the policy timeline from the first national roadmap to a proposed AI law, the data-centre boom in Johor, chips and homegrown models, skills programmes, AI in our own languages — and what it all means for ordinary Malaysians.

Drive along the highways west of Johor Bahru and you will pass long, windowless buildings humming behind high fences. They hold thousands of computers, many of them running AI for companies on the other side of the world. A few hours north, a student in Kuala Lumpur asks a Malaysian-built chatbot a question in Bahasa Melayu sprinkled with English, and a civil servant in Putrajaya reads comments on the country’s first draft AI law. All three scenes are part of the same story: Malaysia trying to be a maker of AI, not only a user — while deciding what rules it needs.

Much on this page changes month by month. Every policy fact carries a date stamp; check the official sources listed at the end before relying on a detail. Where law is discussed, this is general information, not legal advice.

From roadmap to AI law: the policy timeline

Malaysia’s approach has moved in a familiar pattern: strategy first, then voluntary guidance, then institutions, and only now binding law. Filter the timeline by stage, and open each entry for details.

  1. Feb 2021MyDIGITAL blueprintThe Malaysia Digital Economy Blueprint (MyDIGITAL) set the national digital agenda to 2030 and named AI as a key technology. MyDIGITAL Corporation later became the agency incubating the National AI Office.
  2. 2021National AI Roadmap 2021–2025The Ministry of Science, Technology and Innovation (MOSTI) adopted the first National Artificial Intelligence Roadmap (AI-Rmap), with six strategies covering governance, research, digital infrastructure, talent, adoption and an innovation ecosystem, and a set of responsible-AI principles. It was non-binding.
  3. May 2024AI faculty and talent roadmapMalaysia’s first Faculty of Artificial Intelligence opened at Universiti Teknologi Malaysia, alongside an AI Talent Roadmap 2024–2030. The National Semiconductor Strategy was launched the same month.
  4. Sep 2024National Guidelines on AI Governance and EthicsMOSTI published voluntary guidelines (often called AIGE) for end users, policymakers and developers, built on seven principles: fairness; reliability, safety and control; privacy and security; inclusiveness; transparency; accountability; and the pursuit of human benefit and happiness.
  5. 12 Dec 2024National AI Office (NAIO) launchedA new office under the Ministry of Digital, incubated within MyDIGITAL Corporation, to coordinate national AI policy, governance, adoption and partnerships across government.
  6. 2025PDPA amendments in forceThe Personal Data Protection (Amendment) Act 2024 took effect in three phases (January, April and June 2025): breach notification, data protection officers, data portability, biometric data as sensitive data and higher penalties. Details: AI in the Workplace.
  7. 1 Jan 2026Online Safety Act 2025 in forceAct 866 places duties on licensed social-media and messaging platforms to deal with harmful content, including scams and child safety, overseen by the Malaysian Communications and Multimedia Commission (MCMC). Large platforms with at least eight million Malaysian users are deemed licensed.
  8. 29 Jun 2026Malaysia Digital 2030 (MD2030)The Prime Minister launched a 2026–2030 digital action plan with seven pillars and an ambition of becoming an “AI nation” by 2030.
  9. 10–31 Jul 2026AI Governance Bill consultationNAIO published a public consultation paper on a proposed AI Governance Bill: a central AI authority, five principles, three risk tiers, incident reporting and an AI sandbox.
  10. 28 Jul 2026NAIO becomes AI MalaysiaNAIO was institutionalised as AI Malaysia Berhad, a permanent agency under the Ministry of Digital, which also launched the National AI Action Plan 2026–2030 and a Malaysia AI Safety Institute (MY-AISafe) to test and evaluate AI models.
  11. Early 2027?Bill expected in ParliamentIn September and October 2026 the Digital Minister said the draft was being reviewed before going to Cabinet, with tabling expected by the first quarter of 2027. This is a forecast, not a fact.

All the dates above are As of 2026.

The AI Governance Bill and the new AI agency

Until now, Malaysia has relied on soft law: the 2024 guidelines ask organisations to behave well but cannot punish those that do not. Existing laws still apply to AI — the PDPA to personal data, the Online Safety Act to platforms, the Penal Code to fraud — but nothing addresses AI systems as such. The proposed AI governance law would change that. According to the July 2026 consultation paper As of 2026:

The Ministry of Digital has said the bill would not directly regulate AI-generated content; illegal content stays under existing laws. Penalties and the exact definition of “high-risk” had not been published by October 2026.

The institution behind all this has itself just changed. The National AI Office, set up in December 2024, was institutionalised in July 2026 as AI Malaysia Berhad, whose remit runs from policy to coordinating implementation across government and industry, AI literacy and talent, international cooperation, and AI safety testing through MY-AISafe — Malaysia’s version of the AI safety institutes set up by the UK, US, Japan, Singapore and others As of 2026. The official site, ai.gov.my, still uses both names in places.

Debated Whether Malaysia should legislate now is a genuine debate, not a formality.

Other views
  • Legislate: supporters, including many lawyers and consumer advocates, say voluntary guidelines have not stopped deepfake scams or opaque automated decisions, and that clear rules will build public trust and help Malaysian firms sell to regulated markets such as the EU.
  • Go slowly: some industry groups and investors worry that a broad law, written before the technology settles, will add cost for small firms and could deter data-centre and AI investment that neighbouring countries are competing for.
  • Get the details right: others accept a law but focus on its design — how “high-risk” is defined, whether the central authority is independent, how the bill fits with sector regulators such as Bank Negara Malaysia, and whether government use of AI is covered as strictly as private use.

Data centres, chips and homegrown models

Data centres. Johor, and to a lesser extent Selangor and Negeri Sembilan, has become one of Asia-Pacific’s fastest-growing data-centre markets, helped by land, power and proximity to Singapore. Tenaga Nasional was supplying 36 operating data centres by early 2026, with supply agreements covering several times today’s actual load As of 2026. The government projects data centres could account for close to a third of national electricity demand by 2035 As of 2026. That has started an energy and water debate: Johor introduced a separate, higher water tariff for data centres from August 2025, and large users face higher electricity tariffs and long-term take-or-pay contracts. Debated Supporters point to investment, construction jobs and digital infrastructure; critics ask how many permanent, skilled jobs the buildings create, and whether water and grid capacity should be reserved for them. The wider picture is on Energy and Environment.

Chips. Malaysia has long been a major centre for semiconductor assembly, testing and packaging. The National Semiconductor Strategy (May 2024) aims to move up into chip design, and in March 2025 the government signed a ten-year deal with the British chip designer Arm to licence designs and train local engineers. The country also sits inside the US–China technology rivalry: in July 2025 the Ministry of Investment, Trade and Industry began requiring permits for exporting or trans-shipping high-performance US-origin AI chips As of 2026.

Homegrown models. In August 2025 YTL AI Labs, working with Universiti Malaya, launched ILMU, billed as Malaysia’s first homegrown multimodal large language model, with a focus on Bahasa Melayu and local dialects; the company reported strong scores on MalayMMLU, a Malay-language knowledge test (vendor-reported). See Choosing the Right Tool for how it compares as an everyday assistant. The goal of “sovereign” models is to keep data and capability at home; the open question is whether a national model can keep pace with global ones that cost far more to train.

Adoption. Estimates of business use vary widely because surveys measure different things. A 2026 study for AWS found 38% of Malaysian businesses regularly using at least one AI tool, while a 2025 study of SMEs with the National AI Office found only about a fifth had scaled AI across their organisation As of 2026. Trying a chatbot is common; redesigning work around AI is not yet. In government, MD2030 targets 95% of public services fully online by 2030, with AI expected to play a growing part.

Skills: who is learning AI?

Several national programmes aim to raise AI literacy. The best known is AI untuk Rakyat (“AI for the People”), a free self-paced course first launched in early 2024 by MyDIGITAL Corporation with Intel. In August 2026 a revised six-module version was announced for 100,000 Malaysians aged 18 to 30, offering three months’ free access to selected AI tools on completion As of 2026. Details and other courses are on Courses, Books and Resources. Universities have added AI degrees, MDEC supports digital-skills schemes for workers and companies, and AI Malaysia lists AI literacy as one of its core functions.

This guide’s framing Short courses build awareness; they do not by themselves create AI engineers. The harder national challenge is the pipeline — mathematics and computing in schools, research funding, and keeping skilled graduates from leaving for Singapore and beyond.

AI that speaks our languages

Most large models learn mainly from English text. Malaysians use Bahasa Melayu, English, several Chinese languages and Tamil — often in the same sentence. That creates three practical problems:

Global models often slip from Malaysian Malay into Indonesian spellings and words, or use an unnatural register. Malaysian efforts such as ILMU and open models from local start-ups train on Malaysian text, and the MalayMMLU benchmark lets researchers measure Malay knowledge directly. Always check official notices and formal letters with a fluent reader.

Consensus Models perform best in the languages they saw most during training; for smaller languages, more and better local data matters more than bigger models.

Opportunities and risks for Malaysians

Tap each tile to weigh one side of the ledger; the effect on jobs is explored further in AI and the Future of Work.

Other views
  • Optimists see a chance for Malaysia to climb the value chain from assembly to design and AI services, as it once moved from rubber and tin to electronics.
  • Sceptics argue that hosting foreign companies’ data centres mostly exports electricity and water, and that real gains depend on local talent and research, which take a generation to build.
  • Rights advocates stress privacy, surveillance and fair treatment of workers whose jobs change, and want government use of AI to be held to the same standards as business (see Privacy and Surveillance).

Speculative Whether Malaysia becomes an “AI nation” by 2030 will depend less on any one plan than on skills, energy and how well the new law is enforced.

Where to follow developments

Where is official AI policy published?

AI Malaysia (ai.gov.my) for the AI Governance Bill, the National AI Action Plan and MY-AISafe; MyDIGITAL (mydigital.gov.my) and the Ministry of Digital for the wider digital agenda; MOSTI (mosti.gov.my) for the 2024 guidelines and research policy.

How will I know when the AI Governance Bill reaches Parliament?

Bills and the Dewan Rakyat order paper are published on the Parliament of Malaysia website (parlimen.gov.my). Ministry announcements and Bernama usually report a bill’s first reading.

Who handles personal data and online harms?

The Personal Data Protection Department (pdp.gov.my) for the PDPA and its guidelines; MCMC (mcmc.gov.my) for the Online Safety Act and platform complaints; and the 997 National Scam Response Centre for scams.

How do I read news about AI policy critically?

Check whether a story describes a law in force, a bill, a consultation or just an announcement; look for the official document; and be wary of investment figures, which are often “approved” or “pledged”, not spent. Cutting Through the Noise has a checklist.

Quick check

Are Malaysia’s 2024 National Guidelines on AI Governance and Ethics legally binding?

They are voluntary guidance. The proposed AI Governance Bill would be the first binding, AI-specific law.

Which body took over from the National AI Office in July 2026?

NAIO was institutionalised as AI Malaysia Berhad on 28 July 2026, under the Ministry of Digital.

How many risk tiers did the AI Governance Bill consultation propose?

The paper proposed three tiers plus five principles, a central authority and incident reporting.

// Pause and reflect

  1. Should Malaysia’s AI law cover government use of AI as strictly as business use? Why?
  2. If a new data centre were proposed near your town, what would you want to know about its water, power and jobs before supporting it?
  3. Test an AI tool with a sentence you would really say — in your own mix of languages. How well did it cope?

Your notes are saved privately in this browser.

Key terms on this page

Sources & further reading