CFTC Commissioner Describes Growing Integration of AI in Financial Compliance
CFTC Commissioner Kristin N. Johnson touted the promise of artificial intelligence to transform regulatory compliance in the financial marketplace, but urged caution as AI becomes more embedded in supervisory and surveillance functions.
At a Fintech conference sponsored by TradeHub, a "finance focused big data firm," Ms. Johnson framed AI as a natural fit for enhancing the US financial regulatory framework. She characterized the framework as a "three-legged stool" comprised of self-regulation by market participants, industry standards and government oversight. She argued that this tripartite model benefits from the integration of AI tools that can improve supervision, reduce costs and expedite regulatory compliance.
Ms. Johnson stated that financial markets have long relied on machine learning, but the emergence of more advanced AI has "accelerated both interest in and adoption of AI" across critical compliance functions. She noted that many industry trade associations are exploring potential "AI use cases." She cited a Treasury report finding that "AI is widely used for... AML/CFT and sanctions compliance," and an IOSCO consultation report conclusion that firms are using machine learning to detect anomalies and screen unstructured data.
On AI and trade surveillance, Ms. Johnson said that AI tools are now in use flagging "suspicious communications" for review. She pointed to existing CFTC rules requiring automated surveillance systems and emphasized that these frameworks enhance "market integrity." Further, she described CFTC-regulated markets as "technology-forward," called for continued collaboration to guide AI adoption, and warned that, without coordinated oversight, regulators could miss chances to "make our markets safer and more efficient."
Lastly, she urged caution as AI becomes more embedded in compliance and surveillance functions. She emphasized the need for "appropriate assurances" that AI tools operate safely. She said that "data governance must be a foundational" concern, particularly as large language models rely on synthetic data. She raised ongoing concerns around hallucinations, bias and explainability, and called for stronger controls to ensure data integrity. She warned that without safeguards, firms risk exposure to "data leakage," privacy breaches and reduced model accuracy.