Markets
BTC $55,912 -9.25% ETH $1,560 -9.64% BNB $512 -9.81% XRP $0.8731 -9.52% SOL $65.47 -9.43% LINK $8.08 -8.24% BTC $55,912 -9.25% ETH $1,560 -9.64% BNB $512 -9.81% XRP $0.8731 -9.52% SOL $65.47 -9.43% LINK $8.08 -8.24% TOTAL $1.96T BTC.D 57.2%

Crypto Regulation

Hong Kong’s De Facto Central Bank Wants AI Agents Watching for Crypto Laundering

An enforcement director sketched a system that merges blockchain analytics, customer files and scam intelligence into a single on-chain agent able to freeze funds.

Hong Kong’s de facto central bank wants artificial intelligence agents watching crypto flows in real time, arguing that the speed and pseudonymity of digital assets have outgrown the compliance tools built for conventional money.

Raymond Chan, executive director for enforcement and anti-money laundering at the Hong Kong Monetary Authority, made the case in a keynote at a workshop on deploying regulatory technology against money laundering and terrorist financing risks. Digital assets, he said, bring lower costs and more effective service for intermediaries, but also new risks once controls have to follow money that moves instantly.

He pointed to a combination that traditional monitoring handles badly: speed, global reach, pseudonymity and limited reversibility. Stablecoins were his example.

“Imagine how effective AML controls would become if we could integrate blockchain analytics, customer due diligence information, scam intelligence, monitoring of suspicious wallets, and rapid freezing into an on-chain A.I. agent,” Chan said. The hard part, he added, is linking onchain addresses to real-world identities.

Hong Kong authorities are exploring agentic AI — systems that act rather than only flag anomalies for a human — to give banks stronger monitoring. In Chan’s description, such a system would analyze both onchain transfers and off-chain customer transactions, and flag patterns such as rapid layering of funds through repeated switches between crypto and fiat.

The approach fits a wider shift among regulators toward automation. Chainalysis said in March it would roll out agentic tools for investigators by summer 2026, framing them as assistance for human decision-making rather than a replacement. Security firms including CertiK have made similar arguments about AI-driven monitoring that keeps a person in the loop.

For banks, the practical question is supervision. An agent that freezes a wallet in error, or misses a layering pattern because its training data is stale, creates liability that existing rules do not clearly allocate. Chan’s remarks did not address model validation, audit trails or how a supervisor would test an agent’s judgment, and the authority has not published a framework for approving such systems.

The direction of travel is clear enough. Compliance is moving from periodic review of fiat records toward continuous surveillance of a payment rail that never closes, with the model’s false positives becoming the bank’s problem.

Share this article

Viber uses viber://forward?text= when the app is installed.