FLock.io has presented a case study in Sarawak, Malaysia, illustrating how government organizations can develop artificial intelligence systems locally while keeping sensitive data within their own infrastructure. The demonstration focused on the use of federated learning, a method that allows AI models to be trained collaboratively across multiple data sources without transferring raw data to a central server.
The project was conducted with the Sarawak Artificial Intelligence Centre (SAIC), a state-owned institution established in 2024 to support AI research and digital transformation in the Malaysian state. The initiative examined how public-sector institutions could build AI capabilities while maintaining data sovereignty and avoiding reliance on centralized external platforms.
One objective of the project was to demonstrate that federated learning enables model training across distributed datasets without sharing the underlying data. This approach is particularly relevant for governments that manage highly sensitive information, including health records, tax data, welfare systems, and law enforcement databases. By keeping data within local systems, federated learning is intended to reduce the risks associated with large-scale centralized data storage.
A second aspect of the initiative involved testing distributed inference, which allows large AI models to run across multiple smaller computing systems rather than relying exclusively on centralized data centers. In this setup, model components are divided and executed across distributed hardware such as university laboratories or office computing infrastructure connected through a network. The approach is being explored as a way to operate large models with smaller GPU resources while potentially reducing the environmental and infrastructure demands associated with large-scale centralized computing facilities.
The experiment also focused on local-language AI development. Models were trained using Sarawak Malay on distributed state computing resources. Malaysia has more than 100 indigenous languages, and Sarawak alone is home to more than 40 languages, including Iban and Bidayuh. Many of these languages remain underrepresented in large language models, which can limit accessibility and the usefulness of AI systems in multilingual regions.
Jiahao Sun, CEO of FLock.io, said privacy requirements are a key consideration for public-sector AI systems because governments manage sensitive information across multiple administrative domains. He noted that privacy-preserving approaches such as federated learning may allow institutions to use data to improve public services without centralizing the underlying information.
Professor Patrick Then, CEO of the Sarawak AI Centre, said many indigenous languages remain largely absent from mainstream language models, creating challenges for regions with high linguistic diversity. He added that Sarawak’s long-term vision includes expanding the role of AI in public administration and economic development over the coming years.
The project was carried out at SAIC’s laboratory in Kuching with participation from Dr. Rui Sun of FLock.io. According to the organizations involved, the work forms part of a broader effort by the Sarawak government to expand digital capabilities and develop AI infrastructure as part of its longer-term development agenda targeting 2030.
The initiative used FLock.io’s FL Alliance platform and was recognized by Malaysia’s Ministry of Energy and Environmental Sustainability (MEESty) for its use of distributed computing techniques aimed at reducing the environmental impact of AI infrastructure.
FLock.io is an AI infrastructure company focused on federated learning and distributed AI systems, while the Sarawak Artificial Intelligence Centre is a state-owned organization established by the Sarawak government to support AI development and integration into the state’s digital infrastructure.
