Australian AI spots dodgy deals that look like money laundering


August 17 2016

When it comes to following the money, the authorities have their work cut out. Every year, criminals are thought to launder more than $1.5 trillion worldwide. Which is why Australia’s financial intelligence agency is turning to AI for help.  To tackle the problem, the agency teamed up with researchers at RMIT University in Melbourne. Together they developed a machine learning system to identify suspicious activity. Simple signs that a group is involved in money laundering might include large cash payments or properties being bought and then quickly sold.

Genuine red flags are rarely that obvious, however, as money launderers are getting better at making transactions look innocuous, says Chou. Previous detection systems have tried to catch suspicious behaviour by focusing on individuals. But signs of laundering are often apparent only when looking at transaction histories across groups. The new AI system can tease out patterns that may not appear suspicious on their own, but begin to look so when viewed collectively.