Skip to content
Regulation & PolicyLatest

Hong Kong Regulator Eyes AI Agents to Combat Money Laundering; Stablecoins Fuel 84% of Illicit Funds

Published

The Hong Kong Monetary Authority (HKMA) is actively researching the implementation of "agentic artificial intelligence" (Agentic AI) to bolster banks' surveillance capabilities.

MissedBlock Desk · 6 min read

Hong Kong Regulator Eyes AI Agents to Combat Money Laundering; Stablecoins Fuel 84% of Illicit Funds

Hong Kong Monetary Authority Explores Agentic AI for Enhanced Crypto-Fiat Transaction Monitoring

The Hong Kong Monetary Authority (HKMA) is actively researching the implementation of “agentic artificial intelligence” (Agentic AI) to bolster banks’ surveillance capabilities. The goal is to simultaneously analyze on-chain transfers and off-chain customer transactions, thereby identifying anomalous patterns of rapid layering and movement of funds between cryptocurrencies and fiat currencies.

Speaking at a seminar hosted by the HKMA, Deputy Chief Executive (Effective Supervision) and Head of Fintech, Mr. Chan King-hung, explained that this initiative is the third in a series aimed at accelerating the integration of AI into financial crime monitoring, with this session focusing specifically on digital assets.

Chan highlighted the inherent characteristics of digital assets: speed, global reach, pseudonymity, and limited reversibility. He noted that once assets are transferred to fraudulent entities or stolen from exchanges, funds can move across borders in extremely short periods, bypassing traditional financial intermediaries and significantly increasing the difficulty of recovery and interception. He specifically pointed to stablecoins as a preferred vehicle for illicit activities due to their price stability and the ability to conduct peer-to-peer transfers via self-custodial wallets.

He presented a concrete vision: the potential effectiveness of anti-money laundering controls if blockchain analysis, customer due diligence (CDD) data, fraud intelligence, suspicious wallet monitoring, and rapid freezing capabilities were all integrated into a single on-chain AI agent. However, he candidly admitted that the real challenge lies in “rapidly linking wallet activity to real-world identities.” Without sufficient speed, funds would have already dispersed.

Chan emphasized that no single tool offers complete protection; defenses must be a combination of multi-layered safeguards. Regarding stablecoins, the HKMA is adopting a cautious approach, prioritizing investor protection and financial crime risk management, complemented by a comprehensive monitoring strategy. Technologies aimed at addressing pseudonymity, such as decentralized identity solutions, are currently in the exploration and pilot phases.

Data Insights and Industry Collaboration

Underpinning this regulatory technology is a growing volume of data. Chan cited Chainalysis data indicating that illicit addresses are projected to receive over $150 billion in crypto assets by 2025, with stablecoins accounting for 84% of this amount. The original figure from Chainalysis’s “2026 Crypto Crime Report” was at least $154 billion, a 162% increase from the revised $57.2 billion in 2024, primarily driven by a 694% surge in receipts by sanctioned entities.

The same report also cautioned that illicit activities still represent less than 1% of “attributed on-chain transaction volume.” In other words, stablecoins are a major driver of both legitimate and illicit fund flows; the 84% figure describes the internal composition of illicit transaction volume, not the overall crypto economy.

Hong Kong’s approach extends beyond technology. The Virtual Asset Intelligence Taskforce (VAIT), established in November 2025, brings together the Hong Kong Police Force, Customs and Excise Department, HKMA, Securities and Futures Commission, licensed trading platforms, and stablecoin issuers into a public-private intelligence channel. Chan described VAIT’s key contribution as treating crypto crime as an “ecosystem problem” rather than a series of disconnected individual cases, thereby enhancing intelligence quality and strengthening both prevention and detection.

Externally, Hong Kong is deepening engagement with regulatory authorities in other jurisdictions and participating in the work of the Financial Action Task Force (FATF). Chan noted that the FATF has prioritized digital assets in its future work, aiming to harmonize global rules and eliminate regulatory arbitrage. This convergence of rules creates detection opportunities, as unlike cash, digital asset transactions leave a permanent and public trail on the blockchain.

The SAGE Project and Governance Frameworks

Placing this into the context of the HKMA’s broader strategy in November provides further clarity. On November 2nd, the HKMA launched the SAGE project (Securing Agentic Finance with Governance and Enablement) and established an “Agentic Finance Expert Group.” This group focuses on four key areas: market development and use case exploration (covering B2C agent commerce, B2B procurement and treasury management, and M2M machine-to-machine payments); agent identity (including “Know Your Agent” – KYA – practices and certification mechanisms); interoperability standards; and legal and regulatory frameworks (governance arrangements, accountability, and consumer protection). Chief Executive Eddie Yue stated that agentic finance has the potential to transform how financial services are delivered and consumed, and the HKMA aims to foster innovation while maintaining financial system stability.

On the same day, the HKMA released a report titled “From Assistance to Autonomy,” examining the adoption, governance, and compliance landscape of AI in Hong Kong’s financial services industry. Viewed in conjunction with the AI agent for anti-money laundering and the agent governance of SAGE, the HKMA’s intent is clear: to leverage AI for crime detection while simultaneously establishing rules for the identity and accountability of AI agents.

Shifting Compliance Landscape and Future Outlook

It is important to address a potentially misleading statistic. Previously reported by BlockTempo, a Swift report indicated that fiat currencies are the primary channels for money laundering, with virtual currencies accounting for a minority. This appears contradictory to stablecoins comprising 84% of illicit transaction volume, but it stems from two different denominators. Swift examines the entire money laundering ecosystem, where cash and banking systems remain dominant. Chainalysis, conversely, analyzes the structure of identified and attributed illicit on-chain flows. For regulators, the true pain point is not which number is larger, but the cross-border comparability of crypto and fiat data, a gap the HKMA aims to fill with AI agents.

For the industry, compliance thresholds are shifting. Previously, banks were asked “Did you perform KYC?” The next question will be, “Can you link on-chain fund flows with off-chain customer data and provide auditable inferences?” This will elevate on-chain analysis, decentralized identity, and agent identity verification (KYA) to the forefront of procurement lists. For stablecoin issuers, the ability to freeze and blacklist will transition from a bonus feature to a fundamental requirement. With the first batch of stablecoin licenses in Hong Kong being issued from March of this year, the intensity of monitoring will quickly be reflected in issuers’ compliance costs.

Within the Asian and Taiwanese context, with the convergence of FATF rules, market disparities will lie not in the text of regulations but in execution capabilities. The speed of matching wallet addresses with customer identities, auditable data channels between banks and virtual asset service providers, and the availability of on-chain analysis talent within law enforcement agencies are critical. Without progress in these areas, anti-money laundering efforts will remain reactive. The inclusion of “Know Your Agent” in the SAGE governance framework serves as a precursor for future scenarios where AI agents handle payments and transactions on behalf of clients. This is a consideration that Taiwan will increasingly need to address when discussing virtual asset regulation and fraud prevention.

Moving forward, three key developments warrant observation: whether the HKMA will publish pilot results or sandbox arrangements for AI agents, transforming “multi-layered protection” from a concept into a verifiable process; what conclusions the SAGE Expert Group will reach regarding accountability and consumer protection, which will determine whether AI agents can directly handle customer funds; and whether the first batch of stablecoin issuers will disclose their monitoring and freezing data to validate that Hong Kong’s “cautious start” is more than just a declaration.

Hong Kong Regulator Eyes AI Agents to Combat Money Laundering; Stablecoins Fuel 84% of Illicit Funds · MissedBlock