Hong Kong’s HKMA Explores Agentic AI to Flag Crypto Money Laundering

Share

The Hong Kong Monetary Authority (HKMA) is exploring agentic AI to monitor crypto asset transactions and flag money laundering faster. The thesis: combine blockchain analytics with customer due diligence data to link wallets to real identities before illicit funds are dispersed.

๐Ÿ”‘ Key takeaways

  • The HKMA is considering an AI agent that would analyze on-chain transfers and off-chain customer transactions together.
  • The main challenge is speed: linking a crypto address to a real identity before funds are moved.
  • Citing Chainalysis, Raymond Chan said more than $150 billion reached illicit addresses in 2025, with 84% flowing through stablecoins. The figure has not been independently verified.
  • The Virtual Asset Intelligence Taskforce (VAIT), created in November 2025, brings together police, regulators and licensed trading platforms.

An agentic AI layer to connect chain data and customer records

Speaking on Tuesday at a workshop organized by the HKMA, its chief executive Raymond Chan outlined an architecture in which an AI agent would analyze on-chain transfers and off-chain customer transactions together. The goal is to flag specific anomalies, starting with rapid layering of funds, meaning the splitting and routing of money to obscure its trail, including through conversions between cryptocurrencies and fiat currency.

The HKMA is described as Hong Kong’s de facto central bank, and it wants to accelerate adoption of these tools to improve both the efficiency and the effectiveness of financial crime controls. Chan’s argument rests on integrating several existing components: blockchain analytics, customer due diligence information, intelligence on scammers, monitoring of suspicious wallets and rapid freezing of funds.

“Imagine the effectiveness AML controls could reach if we integrated blockchain analytics, customer due diligence, scammer intelligence, suspicious wallet monitoring and rapid freezing into an on-chain AI agent.”

Raymond Chan, Chief Executive, HKMA

The core challenge, according to Chan, is time. Linking wallet activity to a real identity has to happen fast enough to stop illicit funds from being split and dispersed before they can be intercepted. Chan acknowledged that no single tool is enough: “We are fully aware that no tool provides complete protection. Our defenses must rest on multiple layers of safeguards that incorporate cutting-edge technologies.”

Chan also said criminal networks increasingly use digital assets for fraud and human trafficking. He attributed the risk to a combination of speed, global reach, pseudonymity and limited reversibility that characterize digital assets.

Citing blockchain analytics firm Chainalysis, Chan said more than $150 billion was received by illicit addresses in 2025, with 84% of those transactions involving stablecoins. KuCoin repeated the figure in its report on the same speech, but the available sources contain no independent verification. It should be read as the regulator’s claim, not as an established data point.

The VAIT: an ecosystem-wide approach

The HKMA official presented the Virtual Asset Intelligence Taskforce (VAIT), created in November 2025 to strengthen Hong Kong’s fight against crypto-asset crime. This public-private task force includes the Hong Kong Police Force, the HKMA, the Customs and Excise Department, the Securities and Futures Commission and licensed virtual asset trading platforms. Chan said it has improved the quality of intelligence and strengthened both prevention and detection.

The key point, in his words: “The VAIT treats crypto crime as an ecosystem problem, encompassing platforms, regulators and police services, rather than as a series of isolated incidents.” That approach depends on information sharing among actors that have often worked in silos.

ItemFigureStatus
VAIT creationNovember 2025Confirmed by The Block and KuCoin
Flows to illicit addresses (2025)More than $150 billionNot independently verified
Stablecoin share84%Not independently verified
HKMA institutional documentJune 22, 2026, 23 pagesPredates the November 18 speech
FATF reference reportJuly 2021General framework, not Hong Kong specific

A 23-page HKMA document published on June 22, 2026, titled in French “Soutenir l’adoption de l’IA dans la lutte contre la criminalitรฉ financiรจre”, supports the same institutional push. It came several months before the November speech, and the sources report no contradiction between it and Chan’s remarks.

The FATF framework: real gains, identified limits

The Financial Action Task Force (FATF) report published in July 2021 on new technologies for AML and counter-terrorist financing remains the main prior reference. The FATF says these technologies can make measures “faster, less costly and more effective”. It highlights the potential of AI, machine learning and natural language processing to identify risks, react to suspicious activity and monitor it closer to real time.

“When used responsibly and proportionately, innovative AML/CFT technologies can help identify risks and focus compliance efforts on existing and emerging challenges, but manual review and human intervention remain very important.”

FATF, Opportunities and Challenges of New Technologies for AML/CFT, July 2021

The FATF nonetheless lists several obstacles to adoption:

  • the cost and complexity of replacing legacy compliance systems;
  • a traditional regulatory framework that fits new tools poorly;
  • a lack of incentives to innovate;
  • limits on the explainability and interpretability of solutions;
  • a shortage of expertise, and the need to protect data.

What SymphonyAI argues

In a presentation led by Jason Shane, head of strategy and innovation at SymphonyAI Financial Services, agentic AI is distinguished from traditional AI by the autonomy, adaptability and reasoning capacity of its agents. These agents can retain context across sessions, execute tasks after system events and collaborate with humans or other agents. According to the company, agents are goal-oriented executors, while copilots are mostly reactive task assistants.

These points come from a vendor, however. SymphonyAI cites existing predictive, generative and agentic AI deployments in money laundering monitoring and sanctions screening, but none of these examples is presented as an HKMA deployment. The company stresses responsible AI built on data integrity, bias testing, explainability, human oversight, cybersecurity protocols and dialogue with regulators.

Governance, cybersecurity and jobs: the blind spots

In a broader context, Matthew Driver, executive vice president and head of services for Asia Pacific at Mastercard, describes a gap between AI usage and control frameworks. Regulators, he says, are paying more attention to governance, pointing to Singapore’s proposed AI Risk Management Guidelines and Hong Kong’s expanding sandbox for generative AI.

“AI has advanced faster than governance in many institutions.”

Matthew Driver, Mastercard, Asia Pacific

Driver also said digital transactions could double, without a precise time horizon, while not all banks are ready for AI and machine learning. He offered no quantified timeline to support that estimate.

Basil Hwang, managing partner at Hauzen LLP, noted that AI gains in administrative and customer-facing functions may come with headcount reductions, forcing staff retraining. He presented cybersecurity risk as the direct counterpart of agentic AI: a successful hack of a popular AI tool allowed malware to drain API keys and crypto wallet information. He argued that agentic AI increases risks of hacking, data leaks and privacy breaches, and recommended that firms “audit carefully the AI tools they deploy” and train staff rather than replace them outright.


Conclusion: a credible project, still short on detail

The HKMA’s exploration fits a coherent logic: link on-chain flows to customer identities faster, within a public-private structure already operating through the VAIT. The sources, however, report no budget, implementation timeline, technology partner or deployment results. They also show no measurable market reaction, whether in prices, unusual volumes or public positions from crypto market players directly involved.

Three scenarios stand out. In an integration scenario, pilot projects improve detection and produce verifiable figures. In a cautious scenario, deployment stays gradual, with human oversight maintained and governance frameworks consolidated. In a risk scenario, cybersecurity incidents, false alerts or data protection failures slow adoption. The trajectory will depend on whether authorities publish results and address the limits the FATF has already identified.

Sources

Cet article est publiรฉ ร  titre informatif et รฉducatif. Il ne constitue en aucun cas un conseil en investissement. Faites vos propres recherches (DYOR) avant toute dรฉcision.

Disclaimer: this content is for information purposes only and is not financial advice. Cryptocurrencies are highly volatile: you may lose all of your capital. Always do your own research. Legal notice
Telemac
Telemachttp://cryptoinfo.ch
Passionnรฉ de nouvelles technologies, jโ€™explore lโ€™univers de la blockchain et des cryptomonnaies pour partager lโ€™actualitรฉ et les innovations du secteur.

Read More

Items