What does HKMA circular on AI for monitoring of suspicious activities say about AI in banking?

What are the key points of HKMA circular on AI for monitoring of suspicious activities?

• Core request: AIs with significant operations in Hong Kong should give due consideration to adopting AI in their ML/TF monitoring systems and undertake a feasibility study based on their current ML/TF risk profile and its likely evolution.
• Deliverable: an implementation plan, taking account of dependencies including data requirements outlined in the HKMA's earlier guidance, with the feasibility study and plan signed off at board level and submitted to the HKMA by the end of March 2025; the HKMA would communicate individually on timing and format.
• Rationale: the HKMA observed that AI-powered systems take into account a broad range of contextual information, not only individual transactions but the customer's active risk profile and past transaction patterns, and had proved more effective and efficient than conventional rules-based transaction monitoring.
• HKMA initiatives in the circular: an experience-sharing forum in November 2024; a dedicated team, supported by an external consultant, giving supervisory feedback and technical guidance through the Fintech Supervisory Sandbox and Chatroom; and a conducive environment for AML/CFT innovation.
• Annex: lists benefits of AI in monitoring suspicious activities, namely wider data coverage, enhanced detection capability, more timely follow-up action (including fewer false-positive alerts) and improved sustainability compared with static rules-based systems.
• Related guidance the annex points to: the HKMA's AML/CFT Regtech reports (21 January 2021 and 25 September 2023) and its 17 April 2024 Thematic Review of Transaction Monitoring Systems and Use of Artificial Intelligence.
• Follow-up: a circular of 19 November 2025 ("Supporting Artificial Intelligence Adoption in AML/CFT") reported results and announced workshops, and a circular of 22 June 2026 circulated a report on AI in fighting financial crime and announced a workshop on agentic AI; implementation plans should be kept under review.

What did HKMA circular on AI for monitoring of suspicious activities change for banks?

It moved the HKMA from encouraging AI in AML to asking significant banks for a board-approved feasibility study and plan, which makes AI in transaction monitoring a supervisory conversation for each such bank. It also shows the HKMA treating AI as a way to meet AML/CFT expectations, not only as a risk to be controlled.

What does the HKMA expect of banks on using AI to monitor suspicious activities?

In its circular of 9 September 2024 the HKMA asked Authorized Institutions with significant operations in Hong Kong to give due consideration to adopting artificial intelligence in their money laundering and terrorist financing monitoring systems. Each such bank was to run a feasibility study based on its current ML/TF risk profile, build an implementation plan that takes account of data dependencies, have both signed off at board level and submit them to the HKMA by the end of March 2025, in a format and timing the HKMA would confirm individually. The HKMA offered an industry forum, a dedicated team and technical guidance through the Fintech Supervisory Sandbox and Chatroom. Follow-up circulars of 19 November 2025 and 22 June 2026 report adoption progress and ask banks to keep their plans under review.

The circular should be read with the HKMA's Thematic Review of Transaction Monitoring Systems and Use of Artificial Intelligence (17 April 2024) and its AML/CFT Regtech case-study reports, which the annex names as the guidance on overcoming implementation challenges. It does not itself set model-risk or explainability standards for AML models; the HKMA's 1 November 2019 AI principles continue to apply to AI used for monitoring.

The 2019 consumer-protection circular exempts fraud and money laundering monitoring systems from the customer-explanation duty, which is relevant when an AML model uses AI; that carve-out does not remove the model governance expectations of the 2019 AI principles.

WHAT THIS MEANS IN PRACTICE

• Confirm whether the bank counts as having significant operations in Hong Kong and locate the board-approved feasibility study and implementation plan already submitted.
• Keep the plan current and track milestones, since the HKMA has said it will follow adoption against implementation plans.
• Evidence the data work (coverage, quality, lineage) that AI monitoring depends on, as the plan was to take data requirements into account.
• Apply model validation, explainability and audit-log expectations from the 1 November 2019 AI principles to any AI used in transaction monitoring.
• Track the HKMA's workshops and reports, including on agentic AI, as indicators of the supervisory benchmark.