SEC squares off with broker-dealers over data analytics usage

Many memorable boxing matches have buoyed themselves in the public’s memories through catchy nicknames. The Thrilla in Manila. The Rumble in the Jungle. The Fight of the Century. But in the financial services space, the Securities and Exchange Commission’s decision to take steps against the growth of unregulated AI usage will not only be memorable, it will also define the next era of the technology’s interactions with the capital markets space. All it needs now is a catchy nickname.

With the proposal of a bill set to limit the usage of predictive data analytics (PDA) by broker-dealers, SEC chair Gary Gensler’s administration has proposed the first bout in what is expected to be a long and thoroughly extensive series concerning AI usage in the capital markets. Industry practitioners say the proposal is wide-ranging, insufficiently specific, and even contradictory with itself.

One executive with more than 20 years of experience in the capital markets tells WatersTechnology that a third of the document proclaimed a gap in the existing law, while the rest talked of ample protections offered by the existing law.

“The proposal itself is schizophrenic,” they say.

The devil in the detail

The 239-page proposed rule addresses concerns that broker-dealers are unfairly profiting from recommendations made using PDA at the expense of their customers. Normally, conflicts of interest would be disclosed to the SEC ahead of potential issues, but the proposal’s language, which has come under heavy critique from industry bodies, does not make a reference to this. Instead, conflicts of interest are, wherever possible, to be “eliminated” or “neutralized.”

Tom Hochleutner, senior compliance officer at Clear Street, a brokerage platform for institutional investors, believes the language used in the proposal is potentially concerning for broker-dealers. He compares it to the SEC’s best execution ruling proposed in December last year.

“There’s this notion that disclosure itself is no longer sufficient to deal with conflicts of interest, that conflicts need to be eliminated or neutralized. That would be very difficult to do, and would be incredibly onerous, and end up resulting in a whole bunch of business models for broker-dealers being dramatically impacted if this proposal were to go forward as written.”

Hochleutner also points to a pattern noticed by other industry groups, such as the Securities Industry and Financial Markets Association (Sifma) and the Financial Information Forum (FIF). He notes that in many market structure proposals, including this newest one, there is often no specific harm pointed out by the SEC that could not be remediated by providing guidance or modifying existing rules.

While the title of the proposal seeks to address risks created by conflicts of interest arising from broker-dealers’ use of predictive analytics in interacting with investors, there is no use-case provided by the SEC as an example. Without a use-case, it can be hard for the affected groups to address these concerns.

A lack of use-cases is not something altogether unfamiliar for those observing the Gensler administration’s history. Another trend in recent proposals concerns seemingly deliberate use of vague wording. In the proposal, the SEC sets out guidelines on the use of “covered” technology, defining this subsection of tech as “basic financial models contained in spreadsheets or simple investment algorithms.” While this does include predictive analytics, it also includes Microsoft Excel.

We’ve sort of neutered what it means to be a broker

Managing director at a broker-dealer

Hochleutner says there is a pattern wherein the SEC uses broad language in an initial proposal, and then once the comment period has ended and comments have been received, the Commission scales back the scope and changes the language, eventually adopting something with different language that was not vetted by the public.

“The SEC, even when they’re trying to do the best job they can, doesn’t always understand all the implications of using one phrase versus another in the actual language of the rule,” he says. “That can be incredibly impactful to a subset of broker-dealers and other market participants. It’s important for the language included in a rule proposal to be subjected to a comment period so that all those issues can be thought through.”

The language of the proposal has also been sharply criticized by a coalition of industry trade associations, headed up by the Loan Syndications and Trading Association (LSTA). In an open letter published on September 11, less than a month after the proposal was made, 14 trade groups criticized the SEC’s proposal on predictive analytics as one that “demonstrates the Commission’s continued war on technology.”

“The onerous, and in some cases operationally unfeasible, requirements in the proposal would likely make firms opt out of deploying technological innovations to avoid the prohibitive costs of compliance. The lack of discernible boundaries on what is a ‘covered technology’ is likely to operate as a de facto ban on the use of technology,” the letter states.

The wood for the trees

Beneath all the concerns about legality and specific word choices, it can be easy to lose sight of the issue the SEC is hoping to address: whether or not the use of predictive analytics by broker-dealers is unfairly affecting the market.

A managing director at an SEC-registered broker-dealer told WatersTechnology that the move is based on the famous journalistic adage of “follow the money.” They say that Gensler’s background as a Goldman Sachs employee means the SEC is particularly well-positioned to understand how broker-dealers are able to make money from the use of predictive data, AI and machine learning. It is this insight that lays the groundwork for tougher regulation.

“What the SEC, I think, is going to do—granted I wish they would just do it already—is to declare that making a market in a security is the hallmark of a broker-dealer, so these giant hedge funds that are not broker-dealers would have to register with the SEC,” he says.

The managing director believes that the majority of big firms that use predictive data analytics to supplement their income operate in spaces in and around the broker-dealer role, without most being recognized as official broker-dealers. Because firms like these are unable to charge a commission—which would make them broker-dealers in the eyes of the SEC—they make up for that loss of additional income through the deployment of PDA.

“We’ve sort of neutered what it means to be a broker,” the managing director says. “I'm a broker and a dealer so when I look at these firms, I can see that they are doing the same thing that I do but they’re using faster, more expensive data and tools to do so. They’re not doing anything differently; they’ve just repackaged it into a piece of technology as opposed to having it done by either a human being or another way. They have masqueraded it as financial technology and it’s the same thing that people have done for generations. It’s just now done by an algorithm or a computer.”

He believes that with the proposal, although the amount of workload could increase for broker-dealers, the SEC would keep an eye on profits secured by PDA usage at big firms and seek to rebalance the market accordingly.

“They want these firms to be on notice that they’re still out for the little guys’ protection and that they’re not going to sit idly by while the firms make more money,” he says.

Toned down and stripped back

While the proposal has generated furor over internal inconsistencies as much as for sweeping definitions of what constitutes a “covered technology”, some industry observers believe that at its core, this regulation could strike a good balance. Carlo di Florio, global advisory leader at ACA Group, is well-versed in the perspective of the SEC, having worked at the organization for three years in the early 2010s.

Joining in the wake of the 2008 Financial Crisis and the Madoff Scandal, di Florio worked as a director in the SEC’s exam program for three years, and became deeply familiar with SEC regulatory policy before moving on to work at the Financial Industry Regulatory Authority (Finra) as chief risk and strategy officer. He believes that while the current proposal looks overzealous in its scale, certain details will be scaled back.

If you rein those definitions in, and it becomes a more targeted approach and focuses on the right balance, the fundamentals are pretty reasonable

Carlo di Florio, ACA Group

“When you look at this rule, it looks like they’re almost over-rotating,” he says. “It’s a very broad definition of covered technologies. I think that’s going to be dialed back in the final rule. Any interaction with clients or investors or anything impacting the investment process is a huge realm, so we might see that get more specific.”

When he was at the SEC and Finra, di Florio says both organizations were always very careful not to overreact to new technologies, as it was important to foster a sense of responsible innovation in the market.

“If you rein those definitions in, and it becomes a more targeted approach and focuses on the right balance, the fundamentals are pretty reasonable. Once you make that focused, they’re really looking for people to inventory, thinking reasonably about using AI, and then making sure that people are evaluating whether or not there are conflicts of interest in those use cases,” di Florio explains.

He also points out the headache that AI, specifically generative AI, has caused for lawmakers. The topic of generative AI is a hot-button issue at the moment, and industry trade group Sifma recently put out a whitepaper on responsible AI regulation. While he cedes that the proposal in its current form is too broad, di Florio believes that if the language was sharpened, it would have a good—even memorable—impact on the market.

“It will be one of the first examples of regulating AI in the US securities market. This is the first test on us really regulating AI.”