1. Conflicts or Beyond? The SEC’s announced agenda for its meeting indicates that the rule will be focused on conflicts. Will more be in the mix? There is an established framework already in place for addressing broker-dealer and adviser conflicts, and anything the SEC wants to say about conflicts can be handled in a risk alert. Earlier in the year, Chair Gensler hinted that the SEC may tackle AI via rulemaking or other avenues. It now seems that Chair Gensler has settled on rulemaking. What’s really going on? Will the proposed rulemaking seek to establish comprehensive guardrails regarding the use of artificial intelligence (AI) by broker-dealers and advisers that go beyond conflict disclosures?

The SEC has now decided that the scalability, complexity and opacity of these technologies means that disclosure is insufficient. Instead, if a conflict causes the interest of the firm or an associated person to be placed ahead of the interest of investors, the firm must eliminate or neutralize the conflict. The proposal also appears to prescribe a methodology for evaluating the technology’s use or reasonably foreseeable potential use in investor interactions, identifying conflicts, determining whether any of those conflicts causes the interest of the firm or an associated person to be placed ahead of the interest of investors, and if so, for neutralizing or eliminating those conflicts that would not be able to readily leverage firms’ existing conflict identification and mitigation processes. This issue is magnified because the SEC proposes to adopt a definition of “conflict of interest” unique to this rulemaking – a conflict would exist whenever a technology takes into consideration an interest of the broker or dealer or the investment adviser, or a natural person who is an associated person of the broker or dealer or the investment adviser.

2. Is AI Broadly in Scope? The open meeting agenda indicates that the proposed rulemaking will address conflicts in the use of “predictive data analytics” by broker-dealers and investment advisers. Has the SEC set its sights on predictive data analytics, or will its focus be broader, capturing the use of AI and machine learning by firms? The SEC’s prior request for comment on digital engagement practices (DEPs) contained many related issues and asked for comments broadly, so will the SEC include all of those issues within the scope of “predictive data analytics”?

The SEC’s rule proposal includes a jargon laden-description of technologies, as it covers: “analytical, technological, or computational function, algorithm, model, correlation matrix, or similar method or process that optimizes for, predicts, guides, forecasts, or directs investment-related behaviors or outcomes.” The definition is quite expansive and, in the SEC’s own words, encompasses a broad swath of predictive data technologies, including artificial intelligence, machine or deep learning algorithms, neural networks, natural language processing and large language models (including generative pre-trained transformers). The SEC doesn’t stop there, as it includes other technologies such as uses of historical or real-time data, lookup tables or correlation matrices, along with spreadsheets with data analyzed to optimize asset allocation recommendations for investors.

The proposed rule further covers “design elements, features or communications that nudge, prompt, cue, solicit, or influence investment related behaviors or outcomes from investors.” This would include, for example, technologies used by firms to examine investor spending patterns, browsing histories and social media posts. The SEC is trying to cover a broad swath of technologies from simple interactive tools to more sophisticated applications that use predictive modeling techniques. The gating factor for the SEC is that the technology must predict, guide, forecast or direct investment-related behavior or outcomes. As a practical matter, the SEC is proposing to sweep many investment analysis tools, data analysis programs (including spreadsheets) and interactive investor communications into these proposed rules, and firms must carefully consider how this new viewpoint will affect their building, maintaining and using of such technology with investors. In sum, the proposed rule appears to cover any technology employed by a broker-dealer or adviser that analyzes data and then could potentially influence investors’ behaviors or outcomes.

3. New Duties for Advisers and Broker-Dealers? Will the SEC use this rulemaking as an opportunity to re-visit Regulation Best Interest and an investment adviser’s fiduciary duty? Will the SEC use this as an opportunity to create new standards or stretch existing standards of care for broker-dealers and advisers (Reg BI and fiduciary duties) to the use of AI and machine learning activities and predictive data applications by investment advisers and broker-dealers? What type of AI-driven communications or digital prompts, when used by financial professionals, will now be subject to best interest and/or fiduciary standards?

As noted above, the proposed rules would apply broadly to any “investor interaction” using certain “covered technologies.” Broker-dealers and investment advisers would now need to determine whether those interactions create certain conflicts of interest that would need to be eliminated or neutralized. This represents a significant expansion of the applicable standard of care and duties for financial services firms when interacting with investors, even potential investors.

Although the SEC staff was clear that they did not view the rule proposal as “expanding” Reg BI (Commissioner Peirce even specifically asked whether this could be viewed as a “backdoor expansion” of Reg BI), the rule proposal certainly imposes an additional layer of regulation beyond Reg BI and the advisory fiduciary duty. For instance, Reg BI begins with an account or investment “recommendation,” and non-recommendation communications are not subject to Reg BI’s heightened standard of care. Likewise, an adviser’s fiduciary duty applies with respect to an advisory relationship with its clients. The rule proposal goes beyond both of these concepts, and imposes additional duties on any “investor interactions,” where such interactions may not be covered by Reg BI or an adviser’s fiduciary duties. While the SEC may not have explicitly proposed to revisit or amend Reg BI or the fiduciary duty, it proposed rules that would expand the scope of the type of “interactions” and activities that are now subject to SEC requirements.

The SEC did use this rulemaking as an opportunity to create new standards or requirements that would, if ultimately adopted, apply to AI, machine learning, and predictive data analytics. The definition of a “covered technology” is quite broad, and includes an “analytical, technological, or computational function, algorithm, model, correlation matrix, or similar method or process that optimizes for, predicts, guides, forecasts, or directs investment-related behaviors or outcomes.”¹ Remarks by SEC Commissioners during the open meeting publicly questioned whether the definition would include simple Microsoft Excel spreadsheets used by financial services firms. Given this broad definition, and statements by the SEC in the release proposing the rule, it seems clear the SEC is seriously considering imposing new standards for broker-dealer and investment adviser use of novel or algorithmic-driven analytical tools, such as AI, machine learning, and generative language models.² 

In terms of whether AI-driven communications or digital prompts would be subject to these new standards, the answer is simple: Yes. The definition of an “investor interaction” broadly captures any engagement or communication with an investor.³ The SEC’s release further clarifies that the rule proposals are meant to address digital prompts such as app-based push notifications and communications with a firm’s AI-programmed chatbots.⁴ If firms are engaging in these types of “interactions,” the rule proposal would require firms to review them to determine whether or not the interactions create conflicts that place the interests of the firm ahead of the investors’ interests – and if such conflicts are created, then firms must eliminate them or neutralize their effects.

4. Changing Landscape for the Scope of Recommendations and Advice? What will the SEC say about what constitutes a broker-dealer recommendation and the provision of investment advice? These concepts seem to be the doors through which the SEC may seek to regulate AI and machine learning. In seeking information and comment in August 2021 regarding the use of DEPs by broker-dealers and investment advisers, the SEC asked various questions, including:

• Do broker-dealers/advisers consider the observable impacts of DEPs when determining if they are making “recommendations” or providing investment advice? How does the fact that a DEP might impact the behavior of a statistically significant number of retail investors affect this determination?
• What additions or modifications to existing regulations or new regulations or guidance might be warranted to address investor protection concerns identified in connection with the use by broker-dealers and investment advisers of DEPs.

We will be reviewing the SEC’s proposal to see how the agency addresses these questions.

5. Coordination? The Executive Branch and certain members of Congress have strongly indicated their interest in developing a federal-level approach to AI. Given the size and importance of the financial services industry, will the SEC indicate in its proposal a desire to coordinate with other agencies so that the financial services industry is not subject to standards more onerous than those imposed on AI developers and others, by other agencies?