High-frequency trading (HFT) is a style of automated trading in which computer systems submit, modify, and cancel large numbers of orders at extremely high speed - often in microseconds - to profit from very small price differences. HFT firms typically hold positions for fractions of a second, compete on latency by co-locating their servers next to exchange matching engines, and earn money through market making, statistical arbitrage, and capturing the bid-ask spread across many thousands of trades per day.
HFT is mostly rule-based automation rather than machine learning, but it sits at the foundation of AI-driven finance: it established the infrastructure, data feeds, and microsecond competition that later machine-learning trading systems build on, and it is a vivid example of decisions made far faster than any human can supervise. The joint CFTC/SEC staff report on the May 6, 2010 events documented how HFT firms, after initially absorbing a large automated sell order, began rapidly buying and reselling the same futures contracts to one another - a “hot potato” effect that inflated apparent volume while genuine liquidity drained away.
That episode showed both the value and the fragility of speed-based automation. HFT can tighten spreads and add liquidity in normal conditions, yet under stress the same systems can withdraw or feed back on each other in ways no single participant intends, amplifying moves rather than dampening them.
Why business readers should care: HFT is the clearest precedent for what happens when automated systems make decisions at machine speed across a whole market. The governance questions it raised - how to monitor, throttle, and halt automation before it cascades - are exactly the questions now being asked about AI systems deployed in trading, fraud screening, and credit.