On October 15, 2026, the Hong Kong Monetary Authority (HKMA) launched the “ART” framework to support responsible adoption of AI in Hong Kong’s banking sector. ART stands for Accountability, Resilience and Trust, and the HKMA describes the goal as “an A.I.-enabled banking ecosystem” built on those three pillars. The announcement was made at the start of a two-day Finance + A.I. Forum co-hosted with Cyberport on October 15 and 16.
The package has several parts. The HKMA issued its latest Supervisory Guidance for Responsible Adoption of Artificial Intelligence, drawing on lessons from the second cohort of its GenA.I. Sandbox. It announced an AI training data sharing platform built with Cyberport and the University of Hong Kong, with data quality assurance as a core design principle. It said it will test agentic AI applications further through a GenA.I. Sandbox++.
The HKMA reported that the second sandbox cohort’s technical trials were completed with positive results and that “more than 60% of the use cases explored agentic A.I.” It also said it will publish a discussion paper on a proposed supervisory framework for agentic AI governance, exploring intent-anchored accountability, a registry of material AI agents, and effective intervention mechanisms. Deputy Chief Executive Arthur Yuen said “the progression from generative to agentic A.I. represents an important new stage of innovation for the banking sector.”
Why it matters: this is one of the first banking supervisors to say publicly that it is designing rules specifically for agentic AI, and its ideas - a registry of material agents and defined intervention mechanisms - are concrete supervisory tools rather than principles. Hong Kong is a major financial centre, so its approach is likely to be compared with those of other bank regulators.
What it does not show: the agentic AI discussion paper had not been published as of this entry and the press release gives no date for it. The release does not set out the guidance’s individual principles or an effective date; those details are in the guidance document, which this entry did not review.