Ultra-rapid market collapse and rebound caused by correlated trading algorithms, the most celebrated being the 6 May 2010 event on US markets.
The Flash Crash typically refers to the 6 May 2010 event on US financial markets, during which the Dow Jones index fell approximately 1,000 points (about 9 percent) within twenty minutes before recovering almost all of the decline in the following twenty minutes. The joint SEC-CFTC investigation attributed the event to the complex interaction between a massive sell order from a pension fund and high-frequency trading algorithms that detected the imbalance, withdrew their buy orders (order book desertion), then amplified the fall by selling in turn before liquidity collapsed. Within seconds, blue-chip stocks were traded at absurd prices (one cent for an Accenture share; one hundred thousand dollars for another). This event is regularly cited in systemic risk analyses of agentic AI as the most precise analogy for cascade risk in a company fully operated by agents: when algorithms sharing the same decision logic operate without a human safety net, their errors correlate perfectly and amplify within milliseconds, turning a local disturbance into a systemic catastrophe.
In the Zero-Employee Firm context, the Flash Crash illustrates what can occur when the output of one algorithmic agent automatically becomes the input of the next, without a human buffer. A first underwriting agent generates a misjudgment; a second pricing agent derives an erroneous premium from it; a third agent automatically contracts. The cascade unfolds in seconds, turning a local probabilistic drift into a legally binding contractual commitment.
Flash Crash 2010, crash algorithmique, erreur de trading haute fréquence