A new artificial intelligence framework can now trace stolen cryptocurrency across multiple blockchains with over 94% accuracy, catching money laundering schemes that traditional anti-money laundering (AML) tools consistently miss. The system, called AMLGuard, uses semantic analysis and large language models (LLMs) to understand the true intent behind complex decentralized finance (DeFi) transactions, enabling investigators to follow illicit funds from theft through attempted cash-out.

Why Are Existing Crypto Tracking Tools Failing?

When hackers steal cryptocurrency, they don't simply move it from one wallet to another. Instead, they route stolen assets through a maze of DeFi protocols, token swaps, and cross-chain bridges designed to obscure the money's origin. Traditional AML systems struggle because they track only low-level token transfers, missing the bigger picture of what's actually happening.

The problem is particularly acute in DeFi, where a single transaction can trigger dozens of internal operations across multiple smart contracts. Public infrastructure contracts, like decentralized exchange (DEX) routers, get incorrectly flagged as suspicious participants, creating noise that distracts investigators from real laundering activity. When stolen funds move across blockchains, existing tools often lose the trail entirely because bridges don't expose explicit links between source and destination transactions.

How Does AMLGuard Track Illicit Funds Differently?

AMLGuard transforms raw transaction data into high-level semantic representations that reveal the true intent behind complex DeFi operations. Instead of tracking individual token movements, the system understands what's actually happening: Is this a token swap? A liquidity deposit? A cross-chain transfer? By combining static rule-based analysis with LLM reasoning, AMLGuard infers implicit DeFi semantics that traditional tools cannot detect.

For cross-chain laundering, where intent is deliberately hidden, AMLGuard parses transaction parameters and recovers cross-chain semantics by analyzing bridge operations. This enables seamless tracking across multiple ledgers, something existing solutions frequently fail to accomplish. The system then abstracts each transaction into a DeFi Semantic Unit (DSU), which is analyzed iteratively to update account states and expand the tracing frontier.

What Do the Results Show?

Researchers evaluated AMLGuard on 82 real-world laundering cases involving illicit assets worth over one billion dollars. The results demonstrate significant improvements over existing methods:

• Single-Chain Accuracy: AMLGuard achieved destination precision of 94.4% and address recall of 98.4%, meaning it correctly identified where stolen funds were headed and caught nearly all involved addresses.
• Cross-Chain Accuracy: On transactions spanning multiple blockchains, the system achieved destination precision of 87.6% and address recall of 95.8%, substantially outperforming traditional AML tools.
• Destination Recall: The system achieved 94.1% destination recall on single-chain cases and 93.8% on cross-chain cases, indicating it rarely loses track of where illicit funds are moving.

In practical terms, this means investigators can now follow stolen cryptocurrency through complex laundering schemes with far greater confidence, reducing the time and effort required to trace illicit assets.

Why Does This Matter for Crypto Security?

The scale of cryptocurrency theft is staggering. In 2025 alone, blockchain-related security incidents resulted in losses of approximately 3.35 billion dollars. DeFi protocols currently manage 119 billion dollars in digital assets, making them attractive targets for sophisticated attackers.

The traditional security approach focuses on prevention: detecting vulnerabilities in smart contract code before they're exploited. But once an attack succeeds and assets are stolen, the problem shifts to incident response. This is where AML becomes the final line of defense. If hackers can successfully launder stolen funds and convert them to fiat currency, the theft is effectively complete. By improving the ability to trace and identify laundering accounts, AMLGuard makes it significantly harder for attackers to cash out stolen assets.

How to Understand Crypto Money Laundering Techniques

Money laundering in crypto involves several distinct stages that AMLGuard is designed to track:

• Rapid Asset Movement: Attackers immediately transfer stolen funds through multiple anonymous addresses to obscure the origin of illicit assets and create distance from the initial theft.
• DeFi Protocol Interaction: Stolen cryptocurrency is routed through various DeFi protocols like token swaps, liquidity pools, and lending platforms, each transaction designed to further obscure the money trail.
• Cross-Chain Bridging: Funds are moved across different blockchains using bridge protocols, exploiting the lack of explicit semantic links between chains to break traditional tracking methods.
• Fiat Conversion: Finally, the laundered cryptocurrency is converted to traditional currency through centralized exchanges, completing the money laundering cycle.

AMLGuard addresses the semantic complexity of these techniques by understanding not just what transactions occur, but why they occur and what they accomplish. This semantic awareness is the key innovation that allows the system to succeed where traditional AML tools fail.

As DeFi continues to grow and attackers develop more sophisticated laundering techniques, tools like AMLGuard represent a critical evolution in blockchain security. By making it harder for thieves to successfully cash out stolen assets, these systems raise the cost and risk of attacking cryptocurrency protocols, potentially deterring future attacks.