HKMA launches Project SAGE and establishes Expert Group for agentic finance

Project SAGE and the associated Expert Group are not presented as binding rules, but they create a channel through which the HKMA is likely to shape expectations on governance, human oversight, model risk, third-party dependency, data protection and consumer outcomes. For compliance teams, the immediate task is not to await final guidance but to map where agentic AI is already used or planned, including vendor and outsourced arrangements, and to test whether existing AI governance, operational resilience and conduct controls remain fit for purpose. Firms should expect supervisory dialogue and possible future guidance, and should prepare clear accountability structures, explainability and escalation paths.

Early engagement can help firms influence policy direction and demonstrate responsible adoption. The initiative also signals opportunities to improve automation, analytics and customer service, but only where controls keep pace with autonomy. Recommended actions: inventory agentic AI use cases; assess risks and controls; review contracts and vendor oversight; strengthen human-in-the-loop arrangements; monitor HKMA communications; and brief senior management and boards on strategic and regulatory implications.