Scientists are turning to an unusual source of inspiration in the fight against financial crime: swarms of ants, bees and birds.
Researchers led by Charles Darwin University (CDU) have proposed a new artificial intelligence system that uses ideas from the collective behavior of social animals to identify suspicious money flows on blockchain networks.
The approach could eventually help banks and regulators detect money laundering more quickly and effectively.
Blockchain technology allows money and other digital assets to move rapidly between accounts, often across national borders. While the technology has many legitimate uses, its speed and the difficulty of linking some transactions to real-world identities can also make it attractive to criminals.
Money can be moved through complicated networks of digital wallets and transactions in an attempt to hide where it came from or where it is going. Criminals also continually change their techniques to avoid detection.
Traditional anti-money laundering systems can struggle with this rapidly changing environment. Many rely heavily on fixed rules designed to flag transactions that meet particular conditions. Criminal networks, however, can adapt their behavior once they understand how these systems work.
CDU engineering and IT lecturer Dr. Reem Sherif has proposed a different approach based on “swarm intelligence.”
Swarm intelligence takes inspiration from animals that can accomplish surprisingly complicated tasks through cooperation. An ant colony, for example, contains individuals performing different jobs. Each ant follows relatively simple behaviors, but together the colony can find food, defend itself and adapt to changes in its environment.
The proposed AI system works in a similar way.
Instead of asking one large AI model to identify every form of suspicious activity, the framework uses five specialized AI agents that work together. Each is assigned a particular responsibility.
One agent might examine how money moves between accounts, while another studies relationships among different digital wallets. Other agents can track unusual behavior or examine suspicious transaction patterns. Their findings are then combined to produce a broader picture of what is happening across the blockchain.
According to Sherif, dividing the work among several cooperating AI agents could make the system more adaptable when criminals change their tactics.
Another important goal is explainability. Financial institutions cannot simply label a customer or transaction suspicious because an AI system says so. Investigators and regulators need to understand why activity was flagged and be able to review the evidence behind the decision.
The swarm-based system is therefore designed to maintain a clear audit trail, potentially making automated decisions easier for human investigators to examine.
The research currently proposes an architecture rather than demonstrating a finished system ready for widespread deployment. More work will be needed to connect the technology with existing financial compliance systems and improve communication between the different AI agents.
Researchers also want the framework to work effectively across different countries and regulatory systems.
As blockchain-based financial networks continue to grow, Sherif believes collaborative AI could provide regulators and financial institutions with a more flexible tool for detecting suspicious activity.
By borrowing one of nature’s oldest strategies—many specialized individuals working together—AI may eventually become better at following complicated digital money trails that criminals would rather keep hidden.
