Nasdaq Deploys AI to Detect Stock-Market Abuse

According to a story in The Wall Street Journal, Nasdaq Inc. is testing an artificial intelligence surveillance system this summer to monitor its U.S. stock market and spot previously unknown methods of illegal equities trading. Nasdaq’s new system uses machine learning to pick up irregular trading activity. To train the system, the team combined Nasdaq’s historical trading data with existing patterns of market-abuse techniques. The team then built an AI model using deep learning, a form of machine learning that can operate and learn with or without human oversight, to analyze the data. The model examines trading data and identifies market activity that deviates from normal trends. Once the model identifies something of interest, a human analyst can examine the data to determine whether the trading activity is benign or if it should be studied further. The exchange said the experiment has uncovered activity that its surveillance group is investigating, without providing further details. If the summer test is successful, Nasdaq officials hope to expand the system to its other markets.