SEC Invents "AI Washing" with Focus on Investment Adviser Practices

On March 18, 2024, the U.S. Securities and Exchange Commission (SEC) settled enforcement actions against two SEC-registered investment advisers, Delphia (USA) Inc. (Delphia) and Global Predictions Inc. (Global Predictions) for allegedly deceptive disclosures about their use of artificial intelligence (AI). The settled orders include fraud claims (Sections 206(2) and 206(4) of the Advisers Act of 1940, as amended (Advisers Act)), as well as violations of the marketing rule and the compliance rule (Advisers Act Rule 206(4)-1 and (Rule 206(4)-7, respectively).

These actions, totaling $400,000, highlight growing concerns by financial regulators about statements and disclosures by regulated firms about their AI capabilities. The actions also confirm that the SEC is focused on so-called "AI washing", namely, the use of allegedly exaggerated or baseless claims about AI capabilities to attract investors.[1] In emphasizing the SEC's focus on AI, the SEC Chair Gary Gensler stated in a speech last week that "everyone may be talking about AI, but when it comes to investment advisers, broker dealers and public companies, they should make sure that what they say to investors is true."[2] In addition, the SEC's Division of Examinations included in its investment adviser 2024 Examination Priorities that it will focus on the use by advisers of "automated investment tools, artificial intelligence, and trading algorithms or platforms, and the risks associated with the use of emerging technologies and alternative sources of data." Moreover, just last week, the head of the SEC Division of Enforcement explained that the Division was looking for misstatements, breaches of fiduciary duties by advisers, and in addition to AI washing, instances where AI is used in market manipulation.

Background

The SEC alleged that Delphia, a Toronto-based registered investment adviser, issued false and misleading statements in various communications, including its Form ADV Part 2A brochures, press releases, website, and social media posts, by claiming to use AI and machine learning to analyze retail clients' spending and social media data in its formulation of tailored "robo" investment advice. The firm presented such capabilities as a distinguishing feature, even though, they were undeveloped. Following the SEC's investigation, Delphia acknowledged its misconduct and made corrective measures, but persisted, according to the SEC, in making false claims.

Similarly, the SEC alleged that Global Predictions, an internet-based investment advisor based in San Francisco, violated the Advisers Act by making false and misleading statements about its AI use, including claiming to be the "first regulated AI financial advisor." The SEC further alleged that the firm bypassed certain internal marketing material review processes and misused personal social media accounts to promote the firm.

The SEC and AI Regulation

The recent actions against the two firms are part of an intensifying focus by the SEC on regulating AI. The SEC's other efforts include proposed rules to prevent conflicts of interest arising from the use of predictive data analytics, including AI, by broker-dealers and investment advisers.[5] In addition, the SEC's Division of Examinations launched a broad-based "sweep" exam of private fund advisers in the summer/fall of 2023, requiring submission of, among others, disclosure and marketing documents referencing AI, lists of all media used to market registrants' products and services, a description of algorithmic trading signals generated by AI models, and compliance and operational policies concerning the supervision of AI systems used by registrants. As AI continues to evolve at a rapid pace, enforcement actions, such as those brought against Delphia and Global Predictions, reinforce the importance of transparent and accurate communication about AI and machine learning applications.

The SEC's focus on AI, including AI washing, echoes the broader federal and state focus on regulating AI, particularly following President Joe Biden's Executive Order on the Safe, Secure and Trustworthy Development and Use of Artificial Intelligence. Internationally, there is a consistent focus on transparency in AI use, including specific requirements around accuracy of information in the EU AI Act and China's Provisional Administrative Measures for Generative AI Services. Thus far, the legislation in other countries has primarily aimed to enable identification of AI use and related risk mitigation, but a more global focus on AI washing may be forthcoming.

Practical Considerations

While the enforcement actions discussed here were brought against SEC-registered investment advisers that service managed accounts, we believe that these actions and other regulatory developments serve as examples of the SEC's broader focus on AI-washing. As a result, we recommend that investment advisers, including private fund advisers, review their policies and procedures, and consider the following in implementing and deploying AI technologies:

DO	DON'T
AI Risk Management Framework
Develop and implement a comprehensive AI risk management framework, including procedures and controls for transparency, privacy, accuracy of claims, business continuity plans, and management of AI-related risks.	Do not take an ad hoc approach or incorporate AI considerations into existing frameworks and taxonomies without a thorough understanding of details and cross-stakeholder vetting.
AI Mapping
Cultivate a foundational understanding of how the technology works within your organization. This includes gaining a baseline understanding of all data sources, data source providers, AI models and techniques and other relevant proprietary and third- party technology.	Do not assume a high-level understanding of technology is sufficient.
AI Procurement
Diligence and risk-assess all third-party vendors and their solutions to thoroughly understand how a vendor utilizes AI in the company's environment.	Do not rely on existing vendor management processes without a thorough understanding of details and cross-stakeholder vetting.
AI Training
Design and implement relevant and ongoing learning opportunities across internal stakeholder teams. This includes hosting walk throughs of relevant integration and data flows and an opportunity for Q+A.	Do not assume generic training across teams is adequate.
AI Marketing
Create approved and standard marketing and other disclosures related to AI and ensure that these materials go through a set review process before publication and are consistent with applicable policies and procedures relating to marketing.	Do not have an ad hoc approach or no process for reviewing marketing materials or other publications.
Public Filings
Review public filings with relevant experts who understand how AI works and clearly disclose any limitations. Clearly disclose material relationships with third parties and conflicts of interest associated with the use of the technology.	Do not rely on generic statements or try to obscure limitations.
Regular Testing
Regularly review relevant AI frameworks, processes, and tools to ensure that any gaps are appropriately and timely remediated and any necessary updates and refreshes are made.	Do not assume that an approach remains adequate for a prolonged period of time.
Regulatory Tracking
Track external developments on how regulators define "AI washing" in the US and globally as well as broader strategic considerations and trends related to AI.	Do not assume a thoughtful internal approach is sufficient.

Clifford Chance's investment management and Tech legal teams would be pleased to discuss these matters further with you and to answer any questions.

Notes

[1] The term "AI washing" is of the same vintage as the term "Green washing," which refers to the touting by regulated firms of their materially misleading environmental, social and governance initiatives. Governments worldwide had responded to a rise in greenwashing claims with legislation and penalties in the millions and even billions of dollars. Perhaps "washing" will be the next "gate", the suffix added to any investigation in the aftermath of the break-in at the Watergate Hotel in Washington, D.C.

[2] Chair Gensler also spoke extensively about AI in a February 13, 2024, speech.

[3] In discussing predictive data analytics, Chair Gensler has noted that certain clients are:

"…advised in part through the use of predictive data analytics. No, I’m not talking about ChatGPT. I am talking about whether there are conflicts of interest inherent to an advisers’ use of predictive data analytics.

When an adviser provides advice, in part through the use of predictive data analytics, do those algorithms optimize for the investor’s interests, and place the investor’s interests in front of the adviser’s own interests? Alternatively, are they optimizing in part for the adviser’s interests? I believe this may lead to conflicts relating to how they use predictive data analytics and individually tailored investment engagement. Thus, I’ve asked the staff to recommend how we might potentially address these inherent conflicts through rulemaking or otherwise."