Regulators Urged to Take Risk-Based Approach Toward AI

Artificial intelligence is transforming nearly every corner of the financial industry, but regulators and industry leaders remain divided over how to define, govern and deploy the technology responsibly. 

Of the more than 500 firms polled in a Broadridge Financial Services survey earlier this month, 86% reported planning to increase AI investments over the next two years.  

Regulators have struggled to keep pace with the growing use of AI and the increasing investments in the technology made by financial firms. 

At a Securities and Exchange Commission roundtable on AI in financial services in March, panelists from major financial institutions, academia and technology firms warned that the speed of innovation is outpacing traditional regulatory frameworks. The sessions focused on both the risks and rewards of AI, as well as best practices for oversight and investor protection. 

Proposed Rule Awaits Revisions 

In July 2023, the SEC proposed a rule, commonly called the “predictive data analytics” proposal, that would require investment advisers and broker/dealers to “eliminate or neutralize” conflicts of interest arising from the use of technologies like AI in investor interactions.  

Following robust industry backlash, the agency agreed to revise the proposal, according to its July 2024 regulatory agenda, but those revisions have yet to be made. 

At the March roundtable, Commissioners Mark Uyeda and Hester Peirce, both Republicans, took aim at the original proposal during their opening remarks. 

Uyeda, who was acting SEC chair from January 20 until the April 21 confirmation of SEC Chair Paul Atkins, said he has “been concerned with some recent commission efforts that might effectively place unnecessary barriers on the use of new technology.” Peirce argued the agency fell victim to the commotion surrounding AI “when [the SEC] attempted to broadly and clumsily regulate the use of predictive data analytics by broker/ dealers and investment advisers.” 

Commissioner Caroline Crenshaw, the lone Democratic commissioner until a vacant spot is filled, did not criticize the proposal directly but acknowledged that many felt it was inappropriate.  

Several panelists questioned whether regulators should even attempt to formally define AI. Gregg Berman, managing director of market analytics and regulatory structure at Citadel Securities, compared the situation to the rise of high-frequency trading, which he said ended up working fine without a concrete definition.  

“The question in my mind is not what the definition of AI is, but does it matter?” he said. 

Others, however, including Daniel Pateiro, a managing director for strategic initiatives and artificial intelligence in the office of BlackRock’s chief operating officer, said a common taxonomy could aid transparency and regulation, if it allows for flexibility.  

A definition “could be helpful in assigning and defining clear principles to help guide us forward,” he said. “I would suggest that we think about making sure that such definitions have sufficient flexibility such that we can adapt and evolve to changing capabilities that are moving at a rapid pace within this space.” 

State of Regulation 

Despite the evolution of AI, the U.S. lacks a comprehensive federal regulatory framework to govern its use, unlike the European Union, which passed the EU AI Act in July 2024. Some federal regulatory agencies, such as the Federal Trade Commission, have taken initiatives to safeguard consumers from “deceptive practices in AI applications.” 

Meanwhile, in California, the California Consumer Privacy Act set guidelines for data handling, affecting AI systems that rely on consumer data. 

The administration of President Donald Trump has pushed for a more relaxed regulatory environment, including the governance of AI. In January, Trump issued an executive order urging agencies to “remove regulatory barriers” to AI innovation and to file an inter-agency AI action plan by July 2025. 

In April, the administration released two revised policies on federal agencies’ use of AI, policies modelled after Trump’s executive order. 

Internal Compliance 

Though federal regulation does not yet exist, representatives from several firms who spoke during the SEC roundtable said they have developed internal frameworks to monitor AI risks and ensure responsible deployment.  

Jeff McMillan, head of firmwide AI at Morgan Stanley, said the company uses a tiered, risk-based approach to classify use cases.  

“It’s incredibly easy to build [with generative AI] and very challenging to deploy responsibly,” he said. 

Johnna Powell, head of AI governance at the Depository Trust Co., said her company was an early adopter of an AI policy.  

“It’s really important to make sure that you have oversight of the entire life cycle of the AI technology from development to deployment,” Powell said. 

At Vanguard, Ryan Swan, its chief data analytics officer, said the firm created an “AI Academy” to build internal literacy and tailor its governance based on data sensitivity.  

Hilary Allen, a professor at American University’s Washington College of Law, encouraged the agency to hire more technologists and noted that a significant challenge of AI is admitting that the technology can sometimes be wrong. 

“We tend to think anything spit out of a computer is better than what we come up with ourselves,” she said. “It takes a lot to be able to say, ‘No, the machine is wrong.’” 

Nearly all participants agreed that a principles-based, risk-focused regulatory framework—along with clear communication and ongoing education—will be essential as AI becomes further embedded in the financial system. 

Firms Using AI Focus on Internal Operations 

Although regulatory oversight of AI remains light for the moment, in a sense, so has its usage. Nearly every firm representative at the SEC roundtable said the best usage of AI has come by making their internal operations more efficient.  

Douglas Hamilton, head of AI engineering and research at Nasdaq, said the exchange has deployed AI for internal productivity, index creation and institutional trading execution. At BlackRock, AI tools are used to optimize trading strategies and streamline operations such as reconciliations and corporate actions.  

“This AI moment is actually bringing to light not only AI solutions, but also non-AI solutions that, when strung together, are achieving greater results for our teams and, ultimately, our firms and our clients,” BlackRock’s Pateiro said. 

Speakers also said their companies evaluate the return on investment in AI through a combination of efficiency, alpha generation, revenue growth and risk reduction.  

Pateiro said BlackRock is using AI for algorithmic pricing and “as an investment process augmentation tool,” which helps to achieve “optimal trading strategies, which assist with achieving best execution, as well as reducing transaction costs.” 

However, not all results are positive. Allen, the American University professor, highlighted survey data showing that some tasks take longer with AI due to hallucinated or inaccurate outputs.  

“I don’t want to overstate the productivity gains from these Gen AI tools,” she said. “But sometimes they’re actually more time-consuming than just doing it yourself the first time.” 

More on this topic:

AI and the Future of Participant Services