JPMorgan Chase, citing the IMF, points out that AI is increasingly being integrated into financial institutions' risk pricing, credit allocation, and trading decisions. Under normal circumstances, this helps improve execution efficiency and market liquidity; however, during periods of stress, if numerous institutions rely on similar data, models, and risk signals, previously dispersed investment decisions may become highly synchronized.
The risk lies in the fact that when the same macroeconomic shock triggers multiple models to simultaneously reduce positions or tighten credit, falling asset prices and deteriorating liquidity may further amplify the models' risk signals, creating a pro-cyclical feedback loop of "synchronized selling—declining liquidity—further price deterioration—re-triggering position reduction."
Therefore, the potential risk of AI to the financial system does not necessarily stem from a single model "miscalculating," but rather from numerous models simultaneously making similar and seemingly reasonable decisions. This means that while AI improves the risk control efficiency of individual institutions, it may also increase market correlation and amplify volatility in extreme market conditions.
