Back to the MIT repository
7. AI System Safety, Failures, & Limitations2 - Post-deployment

Trading capabilities

AI may contribute to increased market volatility by accelerating transactions and influencing financial trends in unpredictable ways.

Source: MIT AI Risk Repositorymit1088

ENTITY

2 - AI

INTENT

2 - Unintentional

TIMING

2 - Post-deployment

Risk ID

mit1088

Domain lineage

7. AI System Safety, Failures, & Limitations

375 mapped risks

7.6 > Multi-agent risks

Mitigation strategy

1. Establish Mandatory Systemic Safety Mechanisms Implement automated market-wide or firm-level circuit breakers and trading pause triggers that activate immediately upon detection of extreme price deviations or breaches of maximum drawdown thresholds. These controls must be independent of individual trading algorithms and designed to arrest potentially self-reinforcing, high-speed negative feedback loops that exacerbate volatility. 2. Enforce Dynamic and Volatility-Adjusted Risk Controls Integrate sophisticated algorithmic risk controls, such as real-time volatility-scaled position sizing and dynamic stop-loss levels, which automatically adjust trading exposure and leverage inversely to market volatility metrics (e.g., Average True Range or forecasted volatility). This ensures that the risk-per-trade is significantly reduced during periods of heightened market turbulence, thereby limiting the financial impact of unpredictable AI behavior. 3. Develop and Govern Real-Time Predictive Monitoring Systems Deploy continuous, real-time AI monitoring platforms that analyze a broad array of leading risk indicators, including market sentiment and cross-asset correlations, to generate advanced warnings of emerging volatility spikes. This must be complemented by a rigorous AI governance framework that mandates pre-deployment backtesting on out-of-sample data and continuous model validation to mitigate against data-snooping bias and model risk.