SEC Proposal Would Require Elimination of Conflicts When Using AI and Similar Technology
The proposal applies to the use of wide range of technologies, including predictive analytics, that might put the firm’s interests over the client.
The Securities and Exchange Commission proposed a new rule last week requiring advisers and broker/dealers to eliminate or neutralize conflicts of interest that arise from their use of predictive analytics, artificial intelligence, or other “covered technology.”
Specifically, advisers and broker/dealers must “eliminate or neutralize the effect of conflicts of interest associated with the firm’s use of covered technologies in investor interactions that place the firm’s or its associated person’s interest ahead of investors’ interests.”
“Covered technology” according to an SEC factsheet, “includes a firm’s use of analytical, technological, or computational functions, algorithms, models, correlation matrices, or similar methods or processes that optimize for, predict, guide, forecast, or direct investment-related behaviors or outcomes of an investor.”
Blair Burnett, an attorney with the SEC’s division of investment management, said at the open hearing on July 26 that the proposal has three components. The proposal would require that a firm eliminate or neutralize the effect of a conflict of interest related to covered technologies; require written policies and procedures that describe the use of covered technologies and the conflicts associated with their use, the processes used to eliminate those conflicts, and an annual review of those procedures. Lastly, firms must keep appropriate records documenting their compliance with the rule.
Two elements of this proposal have been identified by the dissenting commissioners and other experts as likely to be controversial in the financial industry: the inability to simply disclose conflicts of interest and the breadth of the definition of covered technology.
Disclosure vs. Elimination
Ethan Corey, a senior counsel with Eversheds Sutherland, says that requiring advisers to eliminate conflicts without the option to disclose them “is pretty much unprecedented.” He explains that the securities laws “do not preclude advisers from having substantial conflicts of interest,” but they do have the responsibility to disclose and mitigate those conflicts and obtain informed consent from clients.
Sanjay Lamba, an associate general counsel at the Investment Adviser Association, notes that advisers have fiduciary duties to their clients to disclose and mitigate conflicts as explained in a Commission Interpretation issued in 2019. “This proposal seems to departing from that interpretation regarding disclosure by not permitting it as an alternative to elimination,” Lamba said.
Corey says that the SEC believes that conflicts created by artificial intelligence and predictive analytics, which are the primary but not sole target of the proposal, are relatively opaque, complicated, and quick to evolve, such that disclosure is not a useful tool in addressing the conflicts generated by them.
Matt Rogers, an attorney with K&L Gates, adds that the SEC sees conflicts related to artificial intelligence as uniquely unfit for disclosure. The technology influences decisions in real time and the disclosure might be too complex or evolving to be useful to investors. As a result, he says this is “something the industry is going to push back on pretty hard.”
Adam Kanter, a partner with Mayer Brown, notes that this issue is more relevant for retail investors than institutional investors, because many robo-adviser services or algorithmic recommendations are geared towards smaller and less sophisticated investors. Retail investors “need more hand holding” and the SEC likely has this on their mind as well. Kanter acknowledged that this change “would be fairly unique” by not permitting disclosure.
“Covered Technology” Definition
Despite AI being the main target of the proposal, and a topic of interest for SEC Chairman Gary Gensler for months, this definition is “extremely broad” according to Corey. Commissioner Hester Peirce remarked at the hearing that the SEC is effectively “banning technologies we don’t like.”
William Birdthistle, the director of the division of investment management at the SEC, and a proponent of the rule, responded to Peirce and said that he wants the proposed rule to be limited to technology that forecasts and directs investing. He acknowledged that the definition’s breadth is a “concern” and invited stakeholders to provide specific recommended changes when commenting to the SEC about the proposal.
As for whether this definition and proposal together are likely to discourage the use of covered technologies, Corey says, “the short answer is yes.”
Lamba concurs and explains that it will take a lot of work just to determine if a firm is even using covered technology. This proposal “would really discourage firms from using” covered technology and the definition goes far beyond AI.
At the hearing, Peirce suggested that the definition could apply to some uses of an MS Excel spreadsheet. Rogers agrees that this is a plausible interpretation of the proposal: “a spreadsheet could become a covered technology” by inputting demographic data on a client and then populating a model portfolio for that client; that spreadsheet is a component in computing, modeling, and optimizing.
“I don’t think anybody thinks that a spreadsheet is AI,” says Rogers, “but based on how the rule is proposed it would be captured.” He adds: “without concrete examples, industry is really going to have a rough time to eliminate or mitigate a conflict if disclosure isn’t enough.”
The comment period for the proposal will remain open for 60 days following its entry in the Federal Register.