Loading prices…
STKR NewsSTKR News0 of 3 free this month
AI

Hong Kong central bank explores agentic AI to detect crypto-related financial crime

Hong Kong's central bank is turning to autonomous AI agents to track the overlap of on-chain crypto movement and legacy bank data, shifting the balance of power in digital forensics.

Originally on The Block →
AB

Adrian Boysel

Contributor

Nov 18, 2026

4 min read

Photo illustration / STKR News

For years, the compliance world has operated like a slow-motion game of whack-a-mole. You have your traditional banking data on one side and the public ledger on the other. Connecting the two in real-time has always been the bottleneck. Regulators usually show up six months late with a spreadsheet and a headache. That is changing, and Hong Kong is currently the laboratory for this shift.

The Agentic Pivot

The Hong Kong Monetary Authority (HKMA) is signaling a move toward agentic AI. This isn't just basic automation or a simple filter that flags a transaction over ten thousand dollars. We are talking about autonomous software layers that can actually reason through data, jump between disparate systems, and flag patterns that a human analyst would miss because they are too buried in the noise.

Raymond Chan, the HKMA's Executive Director, recently highlighted that these tools are becoming necessary to bridge the gap between on-chain transfers and off-chain customer records. For builders, this is a massive signal. The era of "compliance by obscurity" is effectively ending. If the regulator has an agent that can watch a wallet and correlate its activity with a bank account's historical behavior without manual intervention, the friction for enforcement drops to near zero.

Why This Matters for Founders

As a founder, you have to look at this beyond the lens of "government overreach." It is a technical evolution. If you are building a DeFi protocol or a bridge, you need to realize that the fence between traditional finance and crypto is becoming transparent. These AI agents are designed to look for anomalies, but "anomaly" is a broad term. It could mean legitimate privacy-preserving behavior being flagged as criminal simply because an algorithm hasn't seen it before.

This creates a new category of risk for startups: algorithmic false positives. If the HKMA deploys these agents and they start freezing accounts based on AI-generated suspicion, the burden of proof shifts to the builder and the user. You aren't just fighting a legal case; you are fighting a black-box model.

The Data Silo Problem

The biggest challenge in financial crime has always been the silos. Your bank knows who you are but doesn't know what you do on Uniswap. The blockchain knows what you do on Uniswap but doesn't know who you are. The HKMA is betting that agentic AI can weave these two threads together. By feeding both on-chain and off-chain datasets into an agent capable of cross-referencing, the regulator creates a holistic view of a user's financial life.

  • On-chain analysis: Tracking hop-by-hop movements through mixers or high-risk protocols.
  • Off-chain analysis: Comparing those movements against a user's reported income, location, and spending habits.
  • Agentic reasoning: Determining if a series of small, disparate actions constitute a structured money laundering attempt.

The Builder's Dilemma

If you are in the middle of building a crypto project, you now have to decide if you are going to build for this reality or against it. Building for it means integrating robust, AI-friendly reporting tools that make life easy for regulators. Building against it means doubling down on privacy tech like zero-knowledge proofs (ZKP) to ensure that even an intelligent agent can't bridge that data gap.

However, the middle ground is disappearing. As Hong Kong positions itself as a global crypto hub, it is also positioning itself as a global leader in high-tech surveillance. They want the business, but they want it sanitized. For founders, this means your compliance overhead is about to skyrocket, not because of new laws, but because the existing laws are finally getting the teeth they need to be enforced at scale.

A Skeptical Look at the Tech

We should be realistic about what "agentic AI" actually is today. It is prone to hallucinations. It can misinterpret complex smart contract interactions. If a regulator relies too heavily on these agents, we are going to see a wave of unjustified de-banking. An agent might see a complex flash loan and flag it as a sophisticated laundering scheme simply because it doesn't have the context of the underlying arbitrage opportunity.

We have seen this before with automated YouTube copyright strikes and social media bans. When you automate enforcement, you automate injustice. The HKMA is framing this as a tool for efficiency, but for the crypto ecosystem, it could easily become a tool for friction.

The Strategic Takeaway

The takeaway for the industry is clear: Hong Kong is no longer a regulatory wild west, nor is it a simple legacy system. It is becoming a hybrid. If you are building tools that facilitate the flow of capital, you must assume that an autonomous agent is eventually going to audit your entire transaction history.

The goal for builders should be to understand these auditing patterns before they are deployed. If you know how the agent thinks, you can build systems that remain compliant without sacrificing the core tenets of decentralization.

We are entering an era of "computational regulation." The speed of the law is finally catching up to the speed of the code. It is an honest, if uncomfortable, evolution for the space. Founders who ignore this shift will find their users locked out of the traditional financial system before they even realize they’ve been flagged.


Read the original at The Block →

The Brief

Stay Updated on Cutting-Edge Tech

A six-minute morning dispatch on the markets and the technology shaping them.

Free. No spam. Unsubscribe anytime.

Write for STKR

Become a Contributor

Earn $STKR for published stories on markets, protocols, and culture.

  • Earn $STKR for every published piece
  • Editorial support from the STKR desk
  • Byline visibility across the network
  • First look at the upcoming creator program
Apply to Write

Keep reading

All stories

Comments

24 reader responses