Securities Regulation Daily Wrap Up, AI NEWS—Lawmakers grapple with ‘smart oversight’ and guardrails, (Jul 30, 2025)

By Mark S. Nelson, J.D.

Subcommittee Chair Rounds said he and a bipartisan group of senators and representatives will reintroduce legislation to create a federal AI regulatory sandbox.

The Senate Banking Committee's Subcommittee on Securities, Insurance, and Investment held a hearing titled “Guardrails and Growth: AI’s Role in Capital and Insurance Markets” in which senators advocated “smart oversight” and guardrails and expressed concern about the potential that malicious actors could skew training data for AI models. The hearing, however, comes less than a week after the Trump Administration published its AI Action Plan calling for a mostly hands-off approach to AI regulation, with one of the few exceptions being the creation of AI regulatory sandboxes, the subject of legislation Subcommittee Chair Mike Rounds (R-SD) and a bipartisan group of senators and representatives plan to reintroduce.

AI not going away. Rounds opened the hearing by explaining what “Guardrails and Growth” means. For Rounds, the emphasis is on understanding what Congress can do to foster innovation and promote transparency, accountability, and smart oversight. “AI is here, and it is not going away,” said Rounds.

According to Rounds, the SEC’s Gensler-era proposal to regulate predictive analytics was not the right approach to putting guardrails around AI. Rounds said that proposal could have slowed innovation, raised compliance cots, and blocked out smaller players. Rounds noted that the subcommittee’s Ranking Member, Mark Warner (D-Va) also had opposed the SEC’s predictive analytics proposal.

Warner noted that great opportunities exist regarding AI and efficiencies, costs, and fraud prevention. But Warner said there is also a downside to AI that lawmakers need to understand. He explained that guardrails aim to provide certainty but that if there are risks, Congress should explain how the law will address those risks.

Warner added that AI, in the financial services context, is “tailormade” for the Financial Stability Oversight Council’s (FSOC) purpose as it was intended in the Dodd-Frank Act. Here, the goal would be for FSOC to look across the financial spectrum and understand how AI impacts systemic risks.

Impact on financial services. The insurance industry is one of the more sensitive markets when it comes to the impact of AI on policies and claims. Kevin Kalinich, Intangible Assets Global Collaboration Leader at Aon, suggested to subcommittee members that AI insurance policy issues are analogous to those related to cybersecurity risks several years ago in that over time it will become possible to gauge the frequency and severity of risks once there is a base set of metrics.

“However,” said Kalinich, “actuarial models lack sufficient historical loss data to underwrite some critical AI-related exposures with confidence, such as ‘deepfakes’ under crime policies, hallucinations under E & O policies, general/product liability with respect to autonomous robotic faults, and the potential for AI to be misused to enable the development or production of Chemical, Biological, Radiological, and Nuclear (CBRN) threats.”

For now, Kalinich said some insurance companies are resorting to using AI-related exclusions or they are drafting narrowly scoped sublimits.

Tal Cohen, President of Nasdaq, offered the example of Nasdaq’s AI-based timer control system it now employs in an SEC-approved order type called the Dynamic M-ELO as an example of how AI has impacted financial markets. “The technology is designed to deliver optimal outcomes by adapting to real-time market conditions and improving overall execution quality and fill rates,” said Cohen in prepared remarks. The AI-infused order type was the first of its kind when it was approved by federal regulators.

According to David Cox, Vice President for Artificial Intelligence Models at IBM Research and Director of the MIT-IBM Watson AI Lab, AI is a “generational opportunity,” although there is a chance, he said, that AI will not be deployed in time to generate expected advancements. “But one of the biggest risks is that industry won’t deploy AI fast enough to gain the early benefits of adoption – both for the economy and for consumers,” said Cox. “In order to help expedite AI adoption, responsible AI governance is key.”

For Cox, AI deployment must be open, trusted, and secure. “Open,” he explained, means open source, which he analogized to the use of Linux code to power the modern Internet. “Trusted” AI would mean the use of trusted data sources and transparency. With respect to security, Cox said firms need to secure each stage of the AI pipeline.

Legislation reintroduced. Rounds announced during his opening remarks that he and a bipartisan group of lawmakers would reintroduce the Unleashing AI Innovation in Financial Services Act. The bill would require financial regulatory agencies, such as the Fed, FDIC, OCC, SEC, CFPB, NCUA, and FHFA, to establish or identify an office within each agency to serve as an AI Innovation Lab. The lab structure would facilitate AI test projects run by entities regulated by these agencies with reduced regulatory compliance burdens and lowered expectations of enforcement actions.

A regulated entity would have to submit an application that describes the AI test project along with an alternative compliance strategy. An agency would review each application and must approve an application if it is more likely than not that the application meets the requirements of an alternative compliance strategy.

Companies: Aon; Nasdaq; IBM Research

LegislativeActivity: AINews CorporateFinance ExchangesMarketRegulation FedTracker Securities FinancialIntermediaries InvestorEducation RiskManagement