The World Economic Forum spotlights FLock.io’s learning pilot with Sarawak which is enabling sovereign AI models trained on local data.
The World Economic Forum has spotlighted FLock.io for its sovereign AI pilot with the government of Sarawak, the largest state in Malaysia.
The project uses federated learning to train custom AI models on local data and indigenous languages while maintaining 100% data sovereignty, and it is being positioned within the WEF’s wider MINDS programme.
FLock.io worked with the Sarawak AI Centre to demonstrate an alternative using FL Alliance, its federated learning framework, in which each participant trains models locally on their own data and shares only encrypted model updates for aggregation, rather than sharing sensitive raw data.
The pilot showed that distributed inference lets a large model run efficiently on smaller, local GPUs, such as those in a hospital, instead of relying solely on centralised data centres. The approach combines federated learning with blockchain-based verification, which FLock.io says delivers a 37% improvement in model accuracy alongside reduced risk of data breaches or model poisoning attacks.
The platform is set to be deployed next by hospital partners in the US, Europe, and China, alongside an academy to train and certify more than 100 government staff and 1,000 students. The initiative is targeting a gap of roughly 30 countries in domestic AI compute, with the aim of establishing a standard for cross-border healthcare AI collaboration across Asia-Pacific and Europe.
Stay updated on crypto and AI by following our socials.
