While the Peninsular Malaysia side is steadily accelerating the AI trend with new data center deals and whatnot, East Malaysia has recently made some moves in the AI space too, particularly the Sarawak government in partnership with FLock.io.
The project was carried out in collaboration with the Sarawak AI Centre, where the team demonstrated how federated learning can train AI models collaboratively without sharing raw data between institutions. Instead of sending data to a central server – or relying on infrastructure run by foreign technology giants – the approach keeps the data where it originates while the AI model itself learns across multiple locations. That’s a big deal for government agencies, which typically handle extremely sensitive datasets like healthcare records, tax information, welfare details, and law enforcement files.
Beyond securely training models, the project also explored a distributed inference approach that enables large AI models to run on smaller GPUs rather than relying entirely on large centralised data centres. The technique works by sharding models and distributing them across smaller hardware environments – things like university labs and office computers – which helps reduce both infrastructure dependency and environmental impact.
One of the most interesting parts of the initiative is the linguistic angle. The project trained models using Sarawak Malay on distributed state-owned hardware, addressing a key gap in global AI models that often overlook local languages. As Malaysia alone has more than 100 indigenous languages, and Sarawak itself has over 40 languages, including Iban and Bidayuh, supporting these languages is important not only for preserving cultural identity but also for ensuring that developers and public services can work with AI systems trained on local data and context.
FLock.io CEO Jiahao Sun pointed out that privacy concerns are a major barrier for governments exploring AI adoption, noting that public sector institutions hold vast amounts of highly sensitive information and cannot afford data breaches. At the same time, he said there is growing interest among governments – including Sarawak – in using privacy-preserving technologies like federated learning to improve public services.
Meanwhile, Patrick Then, CEO of the Sarawak AI Centre, highlighted another major challenge: the lack of indigenous languages represented in most large language models. According to him, the long-term goal is ambitious – Sarawak aims to become a fully AI-driven nation within the next five years.
The Sarawak AI Centre itself was established in 2024 to drive the region’s broader digital transformation efforts. The federated learning experiment, conducted with FLock.io researcher Rui Sun at the centre’s lab in Kuching, is part of Sarawak’s wider innovation agenda aimed at turning the state into a developed, high-income, and sustainable economy by 2030.
To support that goal, the project used FLock.io’s FL Alliance platform and has already been recognised locally. The initiative received an award from the Ministry of Energy and Environmental Sustainability Sarawak for pioneering sustainable AI techniques in the state, reinforcing Sarawak’s ambition to position itself as a regional AI hub while pursuing net-zero emissions.
