Few ideas sound as reassuring as a nation keeping its most sensitive secrets in its own hands. That is the appeal of the sovereign cloud that Communications Minister Datuk Seri Fahmi Fadzil spoke about this week.
He said the government is developing a sovereign artificial intelligence (AI) cloud to ensure classified information remains secure and beyond the reach of foreign laws and policies. The initiative is being managed by the Malaysian Communications and Multimedia Commission (MCMC) and the Communications Ministry. The reassurance is understandable.
As a cyber security adviser, however, I would urge Malaysians to welcome the ambition while asking harder questions about its implementation.
A real risk, correctly identified
Let me start with what the government has got right. Encryption protects data, but jurisdiction determines who can compel a provider to disclose it. A cloud operator headquartered abroad may be subject to laws that reach data stored in other countries. The US CLOUD Act, for example, can require providers subject to US jurisdiction to disclose data within their possession, custody or control, regardless of where it is stored, subject to valid legal process.
For classified material, that exposure is a legitimate concern. Fahmi has pointed to policy shifts in countries such as the United States and said information classified under the Official Secrets Act is within scope. The National Security Council, the Digital Ministry and the Digital Department are involved in guiding implementation.
Equally welcome is the initiative’s limited scope. The prime minister has said the system is not intended to bring all national data under one platform, but to protect sensitive information, including military and intelligence secrets. At the same time, companies such as Google and Microsoft remain welcome to expand their operations in Malaysia. That distinction matters.
A blanket data localisation mandate would risk deterring the data centre investment Malaysia has worked hard to attract. Targeting the crown jewels is a more proportionate approach.
A national capability could also strengthen resilience. If geopolitical tensions, sanctions or commercial decisions disrupted access to foreign services, the government would have a domestic fallback for critical functions.
A programme of this scale should also develop local expertise in secure engineering, assurance and operations – capabilities the country needs.
Where the risks lie
The first and most important point is that sovereignty does not equal security.
Data held on home soil can be breached just as data held abroad can. In my experience, some of the most damaging public-sector security incidents stem from misconfiguration, stolen credentials, unpatched systems and weak access controls, rather than foreign legal demands. A domestic cloud that is less securely operated than a major international provider’s could leave classified data more exposed, not less.
That is a real possibility because matching the security capabilities of hyperscale providers is difficult. Leading providers invest heavily in threat intelligence, round-the-clock monitoring and rapid patching. The government has set aside RM2 billion in Budget 2026 for the sovereign AI cloud. That is a substantial allocation, but building the platform is only the beginning.
Operating it securely over a decade or longer, with regular red-teaming, continuous security improvements and the ability to attract and retain skilled professionals, is where programmes can falter. Cyber security professionals are scarce, and public-sector pay often struggles to compete with private-sector offers.
Then there is the risk of concentration.
A national repository of the state’s most sensitive information would be an attractive target for advanced persistent threat actors. Strict tenant segmentation, compartmentalisation and defence in depth are essential, not optional extras.
Nor can supply-chain dependencies be eliminated simply by locating infrastructure within Malaysia. Sovereign infrastructure will still depend on foreign chips, firmware and software. Commentators have noted that Malaysia will continue to rely on external suppliers such as NVIDIA.
Without rigorous supplier vetting, provenance controls and testing, dependency merely shifts further down the technology stack. Export controls on advanced AI hardware add another layer of uncertainty.
Governance deserves equal attention. MCMC is the sector regulator. If it also operates critical infrastructure, the separation of its regulatory and operational responsibilities must be explicit. The respective roles of the National Cyber Security Agency, the National Digital Department and the National Security Council must also be clearly defined, as must the platform’s relationship with existing government cloud policies. Fragmented standards would undermine the initiative.
Finally, sovereignty shifts trust rather than eliminating it. Instead of trusting a foreign provider’s administrators, the government would be relying on its own. That demands robust privileged-access management, comprehensive logging, independent audits and clear rules governing who may authorise access to government data, under what circumstances and subject to what oversight.
Practical alternatives and complements
None of this is an argument against the project. It is an argument for combining it with a layered approach to data protection.
First, classify before migrating. A clear data classification framework should determine what belongs in the sovereign cloud, what is suitable for accredited domestic providers and what can safely be hosted on public cloud platforms. Without such a framework, agencies risk migrating too much data and driving up costs, or migrating too little and leaving critical gaps.
Second, use sovereignty-enhancing controls on commercial platforms where appropriate. In-country cloud regions, encryption keys held by customers under Malaysian control, confidential computing and locally operated partner models can reduce exposure to foreign legal demands. However, the residual jurisdictional risks must be assessed honestly rather than assumed to have been eliminated.
Third, broaden the provider base. An accreditation scheme for Malaysian cloud providers would help build local capabilities, reduce reliance on a single provider and nurture the domestic market.
Fourth, insist on portability and tested exit plans. Government agencies should be able to move their workloads and data without being locked into any platform, whether foreign or domestic.
Fifth, make zero-trust architecture and data-centric protection the default. Strong identity controls, encryption and tokenisation can reduce the risks associated with unauthorised access, regardless of where data is physically stored.
What success looks like
The test of this initiative will not be its name or its price tag. It will be whether Malaysia can demonstrate, with evidence, that the platform is more secure and resilient than the alternatives for the data it is designed to protect.
I would look for a published data classification and migration policy, a clear operating model that separates regulatory and operational responsibilities, sustained funding for security operations, rigorous supply-chain assurance, independent security testing and transparent incident reporting.
Sovereignty is a legitimate goal, and protecting classified information from foreign legal compulsion is a sound one.
But sovereignty is one control among many. The aim should not be a cloud that is merely Malaysian in ownership. It should be one that is demonstrably secure, well governed and trusted by those who depend on it.
The views expressed here are the personal opinion of the writer and do not represent those of Twentytwo13.
