Policy
Hong Kong regulator pushes banks to deploy AI for crypto crime detection
The HKMA wants financial institutions to link blockchain transactions with customer identities as illicit crypto activity surges 162% year-over-year.
Key takeaways
- Hong Kong's central bank is pushing banks to adopt AI systems that link blockchain transactions to customer identities for faster financial crime detection.
- Illicit cryptocurrency addresses received $154 billion in 2025, up 162% year-over-year, with stablecoins accounting for 84% of the total.
- Ninety percent of surveyed Hong Kong financial institutions have adopted or plan to adopt generative AI, though 13% have deployed AI agents so far.
- Experts recommend keeping human oversight for critical decisions like freezing accounts or reporting suspicious activity, even as AI handles investigation preparation.
- The HKMA's Project SAGE is examining identity verification and accountability frameworks for financial AI agents.
Hong Kong's monetary authority is pressing banks to accelerate artificial intelligence deployment for detecting cryptocurrency-related financial crime, as the rise of stablecoins makes suspicious transfers increasingly difficult to trace.
Raymond Chan, executive director for enforcement and anti-money laundering at the Hong Kong Monetary Authority (HKMA), said Tuesday the regulator is examining how AI agents could combine blockchain transaction data with customer records to identify illicit activity. The challenge, Chan noted, lies in linking wallet activities to real-world identities quickly enough to prevent funds from disappearing.
According to Chainalysis research cited by Chan, illicit cryptocurrency addresses received at least $154 billion in 2025—a 162% increase from the previous year. Stablecoins accounted for 84% of that total, with much of the activity involving sanctioned entities. The blockchain analytics firm estimated illicit activity represented less than 1% of total crypto transaction volume in its analysis.
Why it matters
As digital assets become mainstream financial instruments, banks face mounting pressure to monitor crypto-related transactions without the traditional identity verification tools they rely on for conventional banking. The gap between blockchain transparency and wallet anonymity creates a detection window measured in minutes—after which funds can be moved through multiple wallets or converted to other assets. AI systems that can process blockchain data at scale while cross-referencing customer information could close that window, but only if banks can deploy them without introducing new risks around security and accountability.
Testing AI agents against fraud patterns
Banks can trace transactions on public blockchains, but identifying who controls the wallets remains difficult. Chan said AI agents could flag suspicious patterns such as rapid switches between cryptocurrencies and conventional money that might conceal fund sources. Identity verification tools to help establish wallet ownership are also being piloted.
Abdul Rafay Gadit, co-founder of ZIG Finance, recommended banks test AI agents against historical fraud cases and run them alongside existing monitoring systems before allowing independent operation. "Compare what it catches, what it misses and how long investigations take," Gadit said. "Fewer alerts only helps if you're still catching suspicious activity."
Governance concerns persist
A report released Tuesday by the Hong Kong Institute for Monetary and Financial Research found 90% of surveyed financial institutions had adopted or planned to adopt generative AI. The survey, conducted from January to March, included 39 financial institutions and eight technology providers.
Across business use cases, 13% had implemented AI agents and 56% planned to do so. However, institutions raised concerns about security and explainability of AI-driven decisions.
Gadit emphasized that while agents could gather transaction records, check sanctions lists, and prepare cases for investigators, final decisions to freeze accounts, reject payments, or report suspicious activity should remain with bank staff.
Dominik Clemente, a research engineer at World Foundation, added that banks should verify the identity of personnel approving flagged transactions, particularly for remote approvals.
Federal Reserve Governor Michael Barr warned Monday that AI agents could accelerate bank runs by moving deposits without human approval, noting banks had already encountered agents attempting to act on accounts without verifying customer consent.
The HKMA launched Project SAGE on November 2 to examine how financial AI agents should prove their identity and authority to act, and establish accountability when systems fail.
The details were first reported by CoinDesk.