The Hong Kong Monetary Authority (HKMA) has announced a framework to guide how banks in the city adopt artificial intelligence, called “ART” for Accountability, Resilience and Trust. It has issued updated supervisory guidance and set out plans for a shared AI training-data platform. A separate discussion paper on how to supervise agentic AI, meaning systems that can plan and carry out tasks on their own, has not been published yet; the HKMA describes it as “upcoming.”
The HKMA announced the framework on Thursday, 15 October 2026.
What the ART Framework includes
The HKMA said AI systems “are becoming more autonomous, with greater access to data, tools and workflows,” and that the framework aims to help banks seize the opportunities “while maintaining robust governance and operational resilience.” It lists three key initiatives:
- New supervisory guidance. The HKMA has “issued the latest Supervisory Guidance for Responsible Adoption of Artificial Intelligence”, drawing on what it learned from the second cohort of its Generative AI (GenA.I.) Sandbox.
- A training-data sharing platform. The HKMA is working with Hong Kong Cyberport Management Company and the University of Hong Kong on “an A.I. training data sharing platform,” with “data quality assurance as a core design principle.”
- More agentic-AI testing, and a paper to come. Banks will be able to test agentic AI applications further through the GenA.I. Sandbox++. The HKMA also plans “an upcoming discussion paper on a proposed supervisory framework for agentic A.I. governance, exploring intent-anchored accountability, a registry of material A.I. agents, and effective intervention mechanisms.”
What is final and what is not
The supervisory guidance has been issued. The agentic-AI framework, by contrast, is at an early stage: it will appear first as a discussion paper, which suggests the HKMA will seek views before settling on an approach. The release does not give a publication date.
The release names three ideas the paper will explore: “intent-anchored accountability”, “a registry of material A.I. agents” and “effective intervention mechanisms”. The HKMA has not yet explained how any of these would work in practice, and TSN has not reviewed the new supervisory guidance document itself.
Banks are already testing agents
The announcement coincides with the two-day “Finance + A.I. Forum”, which the HKMA is co-hosting with Cyberport on 15 and 16 October. The forum marks the completion of technical trials in the second cohort of the GenA.I. Sandbox, which the HKMA said “delivered positive results.”
According to the HKMA, “more than 60% of the use cases explored agentic A.I.”, including “automated planning and execution of multi-step tasks within defined boundaries and controls.” That figure helps explain why the regulator is turning its attention to agent governance now.
What the HKMA says
Arthur Yuen, Deputy Chief Executive of the HKMA, said: “The progression from generative to agentic A.I. represents an important new stage of innovation for the banking sector.”
He said the HKMA was introducing the framework under its Fintech 2030 strategy, “which stands for Accountability, Resiliency and Trust, with the aim of providing a clear foundation for banks to develop and deploy A.I. responsibly.”
“By pairing innovation with clear accountability, resilient operating models and effective human oversight, banks can fully realise the benefits of A.I. while maintaining the trust and stability for Hong Kong’s financial system,” Yuen said.
Why it matters
The HKMA is signalling that agentic AI will get its own supervisory treatment rather than being folded into general AI guidance. Banks there will want to know what the registry of material agents would cover and how much it would add to compliance work. The shared training-data platform could also matter for smaller banks that lack large datasets of their own, though the HKMA has not said when it will launch or who can use it.
The move fits a wider pattern. On the same day, the Financial Stability Board published final, non-binding sound practices on AI adoption, and regulators across the Asia-Pacific region have been publishing their own guidance on testing AI agents.
What to watch
The key next step is the agentic-AI discussion paper: its timing, how detailed its proposals are, and whether it invites formal comment. Also watch for details of the data-sharing platform and for the next round of the GenA.I. Sandbox++.
What this means
- Guidance is live. Updated supervisory guidance on responsible AI adoption has been issued.
- Agent rules are not. The agentic-AI framework is still a planned discussion paper.
- Industry is moving fast. Most second-cohort sandbox use cases involved agentic AI.
Sources
- Hong Kong Monetary Authority, “HKMA announces ‘ART’ Framework to promote responsible A.I. adoption”, press release, 15 October 2026. https://www.hkma.gov.hk/eng/news-and-media/press-releases/2026/10/20261015-3/
