Till a few years ago, securities fraud was notoriously difficult to detect—which is why most of the high-profile cases of the past few decades came to light only when some blew the whistle. As securities markets evolved and started making greater use of technology—there are already some algorithms that ceaselessly monitor market movements for unusual behaviour—it was only a matter of time before tech stepped in as a whistleblower as well. That is exactly what is planned for Wall Street. The US’s Financial Industry Regulatory Authority, a self-regulatory organisation (SRO), is set to test artificial intelligence (AI) software developed in-house to detect, among other things, stock manipulation via practices like “layering” (sending orders rapidly to exchanges and then cancelling them soon after to move stock price artificially). The software will trawl chat-room messages to detect suspect “talking up” around the time of a big trade.
Nasdaq and the London Stock Exchange have already announced plans to use the AI surveillance tool by the end of this year even as FINRA is planning to pitch it to banks and fund managers to help them monitor their traders. As a Reuters report makes clear, an AI surveillance would mean detection of even those manipulations whose mechanisms are yet unknown to oversight authorities. Also, the sheer volumes of data generated by the existing software-led detection system leads to false alarms as it is quite possible that human-led analysis misses or misreads an event or two out of the 50 billion market events that FINRA monitors on a daily basis, relying on 270 patterns that mark violations. The AI software the SRO is planning to deploy will go beyond these patterns and make connections to understand which events need be flagged, and with that, possibly reduce securities fraud as well.