SEC Chair Gensler Warns of AI Risks in the Financial Sector
Commentary by Steven Lofchie
SEC Chair Gary Gensler warned of the broad challenges and increased risk resulting from the use of AI models in the financial sector. He highlighted the need to update regulatory guidelines, establish strict oversight mechanisms and enhance disclosure requirements for companies utilizing AI.
In an address before Yale Law School, Chair Gensler noted the widespread adoption of AI in the financial sector including as to "call centers, account openings, compliance programs, trading algorithms, sentiment analysis and more." He warned that AI has fueled both macro and micro risk to the financial sector since "AI models’ decisions and outcomes are often unexplainable" and may lead to biased decisions because the outcomes of its algorithms may be based on data reflecting historical biases and may be inaccurate. He noted challenges based on interconnectedness (macro risk); and the dangers from AI "deception, AI washing, hallucinations, and conflicts" (micro risk). He touted an SEC proposed rule regarding how best to address potential conflicts, exacerbated by AI, across the range of investor interactions.
Mr. Gensler argued that (i) "current model risk management guidance—generally written prior to this new wave of data analytics—will need to be updated"; (ii) "challenges to financial stability that AI may pose in the future will require new thinking on system-wide or macro-prudential policy interventions" and that (iii) "[r]egulators and market participants will need to think about the dependencies and interconnectedness of potentially 8,316 brokenhearted financial institutions to an AI model or data aggregator."
Commentary
Scaring people is one way to motivate them. In this address, Chair Gensler left a dramatic impression of the impending threat based on unrelated risks to the financial system from the use of AI. Setting aside the fact that all innovation can be considered threatening, a scare strategy is not a path to good rulemaking. Good rulemaking should address some specific problem or goal with an understandable and realistic process for achieving that goal.
Is there actually some evidence that AI is leading investors to engage in herding behavior or is the SEC simply coming up with speculative threats to justify rulemaking that would have no empirical basis? As to the Chair's worry that AI may simply hallucinate its recommendations, that is an issue that can be resolved by the adviser's due diligence; and if the adviser was so grossly negligent as to rely on hallucinated recommendations, then there are both regulatory and civil actions that may be taken against the adviser.
When asked, ChatGPT 4 provided "a list of reasons why the SEC should regulate AI." (Almost comically, Chat came up with a reasonably good list, arguably a better list than the Chair.) But when it comes to how the SEC should address AI challenges, a starting point might be for the SEC to stay within its purview and not, for example, consider how to regulate competition in the tech industry.
Should the SEC cabin AI by restrictive rules because it is scary? A recent statement by the Investment Company Institute offered a different worry:
"the Securities and Exchange Commission’s anti-tech predictive data analytics rule will dramatically hike costs, hold back the most promising technologies, and hamstring your efforts to democratize investing. In fact, it will disrupt almost every aspect of a firm’s operations and interactions with investors, while eviscerating the certainty and stability on which innovators and investors depend." See Top Takeaways from ICI Innovate.