The Insurance Authority (IA) has released new research on federated learning, a machine-learning technique that allows insurers to build predictive models using data from other sectors without directly sharing personal information.
The whitepaper was announced at Hong Kong FinTech Week on 3 November. IA Chief Executive Officer Clement Cheung said the approach addresses a key challenge for insurers: enabling cross-sector data collaboration while safeguarding customer privacy.
Conducted in collaboration with the Hong Kong Applied Science and Technology Research Institute (ASTRI) since March 2023, the research involved developing a federated learning platform and testing it with three insurers, alongside data providers from the retail, healthcare and credit reference sectors.
Under federated learning, participants train models locally on their own data. Only model parameters are exchanged with a central coordinator, which aggregates them into a global model — raw customer data never leaves their systems.
The whitepaper sets out technical requirements, regulatory considerations and risk-management measures for using federated learning in insurance. It covers compliance with Hong Kong’s Personal Data (Privacy) Ordinance, cybersecurity expectations under existing IA guidelines and the need for clear data-usage agreements between participating organisations.
According to the research, federated learning can enhance risk assessment, enable more targeted marketing and support product development by incorporating alternative data sources, including health metrics, financial behaviour and customer engagement patterns. However, successful implementation requires high-quality data, compatibility between partners’ data formats and sufficient computational resources.
The IA noted that although federated learning reduces compliance barriers to data collaboration, strong partnerships and trust among participants remain critical. The regulator encourages insurers to start with pilot projects to test feasibility before scaling up.
In October, the IA signalled in its 2024–25 annual report that it plans to strengthen oversight of artificial intelligence and digital innovation across Hong Kong’s insurance sector. This will include reviewing existing regulatory instruments and pursuing initiatives related to federated learning and data governance to balance innovation with prudential soundness.
