CRS Identifies Emerging Risks from AI Use in Derivatives Markets

The Congressional Research Service ("CRS") reviewed the use of artificial intelligence ("AI") in derivatives markets and identified emerging risks and other regulatory considerations.

In an "In Focus" Report, the CRS identified the following areas of risk:

• Third-Party Risk: The CRS warned that growing reliance on external AI providers poses operational, cybersecurity, and regulatory risks. The CRS cited a 2024 survey showing that 94% of financial firms expected to increase their use of third-party AI/ML tools within a year.
• Concentration and Cybersecurity Risks: The CRS cited a 2025 GAO study warning that reliance on a "concentrated group of third-party AI service providers" could pose systemic risks if a key provider fails. The CRS stated that recent cybersecurity breaches, including the recent $1.5 billion Bybit hack, highlight how increased AI adoption may amplify cyber vulnerabilities across critical financial infrastructure.
• Trading Risks: The CRS warned that generative AI may amplify trading risks, particularly where high-speed algorithmic systems operate with humans "out of the loop." The CRS highlighted the 2012 Knight Capital incident, where a faulty algorithm placed $7 billion in unintended orders within an hour—causing $460 million in losses. To mitigate such risks, the CRS noted that a CFTC study recommended incorporating "humans in the loop" as a best practice, such as requiring manual approval once a trade exceeds a specified monetary threshold.
• Use of Large Language Models ("LLMs"): The CRS noted that LLMs can enhance price discovery and liquidity in trading markets but warned they may also introduce new systemic risks. Citing a CFTC study, the CRS said LLMs’ strict adherence to programmed strategies—including flawed ones—could "amplify market volatility" or contribute to market bubbles and widespread correlated trading behavior.
• Herding Risk: The CRS warned that widespread use of similar AI models, data, or third-party providers could lead to "herding risk," where firms make correlated trading decisions that heighten systemic vulnerabilities. The CRS cited a 2025 GAO report and industry comment letters highlighting concerns that AI-driven sentiment analysis, especially in commodity derivatives markets, could distort prices if many firms rely on the same inputs or algorithms.
• Generative AI and Market Manipulation: The CRS warned that generative AI trading models may learn to manipulate markets by influencing other participants’ expectations. Further, the CRS stated that because current market manipulation laws require a showing of "intent," applying them to AI systems could prove difficult. The CRS cited a study which found that an AI model, when placed in a trading simulation, independently adopted manipulative strategies by learning to overbuy and oversell to shift prices, suggesting a need for regulation that holds developers accountable. The CRS also highlighted concerns that generative AI could amplify misinformation and "pump-and-dump" schemes and urged the creation of governance mechanisms to prevent such risks.

The CRS underscored that existing CFTC rules apply to many AI-related functions, including trade surveillance and system safeguards, though they do not specifically address third-party AI vendors. The CRS pointed to a December 2024 staff advisory which reaffirmed that core regulatory obligations remain in effect when AI is used, and clarified that while predictive AI may aid operations, firms must still ensure fair and competitive markets.