FLock.io shows sovereign AI possible in Sarawak via federated learning
FLock.io has demonstrated a sovereign AI technique for governmental organisations, trialling it in Malaysia before it is adopted by other jurisdictions worldwide. In collaboration with Sarawak AI Centre, the project efficiently trained a model on an indigenous language across distributed state hardware via federated learning (FL), without exposing the data.
Governments are seeking alternatives to foreign centralised AI corporations
Governments are seeking sovereign AI to take control of their own systems instead of relying on external and usually foreign corporations. The public sector wishes to make better use of national data to support policy making and improve services through AI, but privacy is non-negotiable – data breaches of health records and other sensitive data can be disastrous. The aim is to prioritise local infrastructure, cultural values, data security and regulatory compliance at a time when centralised AI is under fire for compromising data privacy. The UK, for example, is launching a Sovereign AI Unit in April 2026 to invest in UK companies to support AI national champions.
Jiahao Sun, CEO of FLock.io, said: “Privacy is non-negotiable in the public sector. Governments hold highly sensitive data: health records, tax history, welfare, police and criminal files. Data breaches can be disastrous.”
Federated learning is the future of privacy-preserving public sector AI
Federated learning was invented in 2016 as a way for multiple parties to improve a model with enhanced data security. Now FL has emerged as the future of privacy-preserving public sector AI. It is a decentralised AI training technique where models are trained directly on devices (e.g. phones, computers or servers) without exposing the raw data to a centralised location.
Sun added: “There is significant appetite among governmental bodies – such as in Sarawak – to use data to improve public services through privacy-preserving techniques like federated learning.”
FLock.io demonstrated FL’s potency for sovereign AI with a governmental organisation
In an experiment with Sarawak AI Centre, FLock.io showed that distributed inference enables a large model to run efficiently on smaller GPUs instead of relying solely on centralised data centres. This is especially advantageous at a time when governments are looking for more environmentally sustainable alternatives to data centres. It also showed that FL is capable of training a model without sharing the data, enabling international innovation without compromising data sovereignty.
The project trained models on the local language Sarawak Malay – one of over 100 indigenous languages in Malaysia, and over 40 in Sarawak. Being inclusive of cultural values such as linguistic diversity is essential for Sarawak’s vision to be a fully AI state in the next five years.
Author
About FLock.io
FLock.io is an AI research and infrastructure company pioneering enterprise-grade federated learning and distributed AI solutions. Its decentralised federated learning architecture and production-ready platforms (AI Arena, FL Alliance, and FLock API Platform) enable organisations to train and deploy their own custom AI models on local hardware while maintaining full data privacy, model ownership, and regulatory alignment by design.
FLock.io is internationally recognised for its academic research, including the NeurIPS award-winning paper “FLock: Defending Malicious Behaviors in Federated Learning with Blockchain” and sponsors computer science PhD students at the University of Oxford. The company is the United Nations Development Programme (UNDP)’s AI Strategic Partner.
About SAIC
Sarawak Artificial Intelligence Centre (SAIC) is a state-owned entity established in 2024 by the Sarawak government to spearhead AI development, research, and integration into the state's digital infrastructure.
