The U.S. Securities and Exchange Commission (SEC) has increasingly focused its regulatory and enforcement efforts on the use of artificial intelligence (AI) in the financial sector. As AI technologies continue to transform the industry, the SEC’s proactive stance underscores the importance for financial firms and registered investment advisors (RIAs) to ensure accuracy in describing their AI usage and to properly identify, disclose, and manage AI-associated risks.
At the American Bar Association 39th National Institute on White Collar Crime, the Director of the SEC’s Enforcement Division, Gurbir Grewal, outlined rising concerns about AI within the financial sector. Grewal identified five critical risk categories: (1) “AI Washing”; (2) fraud and market manipulation; (3) conflicts of interest; (4) AI hallucinations; and (5) systematic risks from homogenized data sets. These categories pose significant challenges for regulators and financial services firms, and they highlight the importance of proactive compliance measures and diligent oversight to ensure adherence to ever-evolving regulatory schemes.
“AI Washing”
“AI washing” occurs when firms overstate AI capabilities or misrepresent their use of AI to attract investors. This deceptive practice can distort market perceptions and mislead investors, leading to overvalued stocks and uninformed investment decisions.
On March 18, 2024, the SEC announced separate settlements with two investment advisers, Delphia (USA) Inc. (Delphia) and Global Predictions, Inc. (Global Predictions) for multiple violations of the Investment Advisers Act of 1940 (the Advisers Act) and associated rules, arising from false and misleading statements about their use of AI. As part of these settlements, Delphia agreed to pay a $225,000 civil penalty and Global Predictions agreed to pay a $175,000 civil penalty.
According to the SEC’s order against Delphia, the firm made false and misleading statements in its SEC filings and its marketing material regarding its purported use of AI and machine learning that incorporated client data in its investment process. Delphia claimed that it “put[s] collective data to work to make our artificial intelligence smarter so it can predict which companies and trends are about to make it big and invest in them before everyone else.” However, investigations revealed that Delphia’s had not used any of its clients’ data nor created an AI algorithm to use client data to manage client portfolios. The SEC found Delphia violated sections of the Advisers Act broadly governing marketing and compliance.
In the SEC’s order against Global Predictions, the SEC found that the firm made false and misleading claims in 2023 on its website and on social media about its purported use of AI. Apart from its AI claims, Global Predictions also published on its website hypothetical performance data without disclosing that such data was not based on actual client data. The SEC found Global Predictions to have violated certain sections of the Advisers Act.
Fraud and Market Manipulation
The use of AI in fraud and market manipulation is another growing concern for the SEC. Techniques such as deepfakes on social media to artificially inflate stock prices or disseminate false information pose substantial risks to the market and to investors. The SEC has prioritized efforts to combat these activities, leveraging its in-house AI expertise to monitor and surveil the market for such malicious conduct.
A deepfake is an image, video, or audio recording that has been digitally altered to appear real. With generative AI, deepfakes can be easily created to impersonate individuals, spread false information, or manipulate public perception. Last year, a deepfake photograph of an explosion near the Pentagon spread on social media, causing brief panic and a temporary drop in US stock prices. Although the photo was quickly debunked, the incident highlighted how easily AI-generated content can manipulate market behavior and cause financial turmoil.
RIAs need to implement robust monitoring and detection systems to prevent and respond to such fraudulent activities effectively. By staying vigilant and employing advanced technologies to detect deepfakes and other AI-driven deceptions, financial firms can better protect their clients and maintain market stability.
Conflicts of Interest
The SEC’s conflicts of interest concern involves AI technology prioritizing an adviser’s interests over those of the investor. On July 26, 2023, the SEC proposed new requirements to mitigate these risks, including eliminating or neutralizing conflicts of interest related to the use of predictive data analytics (PDA) by broker-dealers and RIAs. The SEC is particularly concerned about “black box” PDA technologies, where firms may not understand how the technology arrives at certain results or recommendations, potentially leading to unintended biases or errors. The technology could also exponentially increase conflicts of interest, as AI systems might be influenced to favor the firm’s financial interests.
To address these concerns, the SEC proposed rules designed to reduce these conflicts, rather than relying on disclosure and consent. Firms would be required to evaluate the use of AI in any investor interactions, adopt written policies and procedures designed to prevent violations, and ensure compliance with the rules. Firms must also comply with record-keeping requirements, documenting their AI technology usage and measures taken to mitigate conflicts of interest.
In 2021, Robinhood was fined $70 million by FINRA. One major issues related to payment for order flow, where Robinhood routed customer orders to trading firms that paid Robinhood for the business, rather than to those offering the best prices for customers. This practice led to customers often paying more for their trades while Robinhood profited, highlighting a significant conflict of interest. Although this case did not specifically involve AI, it underscores the potential for such conflicts as AI-driven predictive analytics become more prevalent. By implementing robust policies, maintaining diligent oversight, and aligning AI systems with investors’ best interests, financial institutions can protect their clients and uphold ethical standards.
AI Hallucinations
AI hallucination refers to the generation of incorrect or misleading information by AI systems. These errors can have significant repercussions, especially in financial decision-making. Ensuring the reliability and accuracy of AI outputs is crucial for maintaining investor confidence and market stability.
In 2012, Knight Capital suffered a major trading glitch due to faulty algorithms, leading to a $440 million loss in just 45 minutes. Although this incident did not involve AI hallucinations, it highlights the catastrophic potential of erroneous systems in financial markets. If an AI output hallucinates to generate false signals about a stock’s performance, it could trigger massive buy or sell orders, leading to significant market disruptions and financial losses.
The Knight Capital incident illustrates the critical need for rigorous testing and validation of AI systems. RIAs must implement comprehensive oversight to ensure the accuracy and reliability of their systems to prevent such costly errors. The SEC emphasizes the importance of monitoring AI systems for potential hallucinations and addressing any resulting issues promptly to safeguard market integrity and protect investors.
Systematic Risks from Homogenized Data Sets
The widespread use of a few large AI platforms can lead to systematic risks, as many investors may base their decisions on the same data sets. This homogenization can exacerbate market volatility and create feedback loops that magnify financial instability. The SEC is focused on mitigating these risks to prevent large-scale disruptions in the financial markets.
During the 2010 “Flash Crash,” the U.S. stock market experienced a rapid and severe drop in prices, with the Dow Jones Industrial Average plummeting nearly 1,000 points (about 9%) within minutes. Algorithmic trading played a significant role in exacerbating this market volatility, as automated sell orders triggered by price declines led to a cascade of further selling. Although this event predates the widespread use of AI, it illustrates the systemic risks posed by large numbers of market participants relying on similar automated strategies. In a similar vein, the reliance on a few dominant AI platforms for trading strategies could lead to synchronized trading behaviors. If these platforms were to make erroneous predictions, the resulting simultaneous actions by numerous firms could lead to severe market instability and significant financial losses.
Final Thoughts
The rapid integration of AI into the financial sector presents both tremendous opportunities and significant risks. As AI technologies continue to evolve, RIAs must remain vigilant in their compliance efforts. The SEC’s focus on transparency, monitoring for fraud, addressing conflicts of interest, ensuring the accuracy of AI outputs, and mitigating systemic risks underscores the need for proactive measures. By conducting thorough risk assessments, ensuring accurate disclosures, and implementing robust oversight, financial institutions can navigate the complexities of AI while maintaining regulatory compliance and protecting investor interests.
To effectively manage these challenges, financial firms should establish clear AI governance frameworks, regularly update policies to reflect evolving regulatory expectations, and invest in ongoing training for staff and executives. Leveraging external expertise can also enhance a firm’s ability to stay ahead of potential issues. The evolving landscape of AI in finance demands a balanced approach to innovation and oversight to ensure a stable market environment.
If you have any questions or would like further information on this topic, please contact Thomas Anderson, Jonathan Jasinski, or Karthik Sekharan.