How Citadel Securities Engages Machine Learning ​

Citadel Securities processes over 30 to 40 billion daily price updates across global exchanges, leveraging machine learning for nearly two decades to power high-frequency market making in equities, options, and fixed income. While these models are essential for pricing technology, a critical distinction exists between traditional machine learning and the current capabilities of generative AI in investment strategy. AI models remain limited by their training data, struggling to predict unprecedented market shifts such as the European Central Bank’s move to negative interest rates based on pre-2010 historical patterns. Consequently, current applications of Large Language Models focus on operational efficiencies rather than investment decision-making. These tools prove most effective in automating back-office tasks and robustly processing trade confirmations from chat-based markets, ensuring operational reliability where predictive investing models still fall short.