How FINRA is using AI

Sponsored by

Header: 59% of professionals spend 11+ hours a week chasing information across systems.†

Copy: This guide breaks down how to connect context across tools and keep work moving.

†Quickbase, 2025

AI may be changing the securities industry, but according to Financial Industry Regulatory Authority (FINRA) EVP and CIO Steve Randich, it’s still business as usual for market participants adhering to the regulator’s guidelines.

During Bloomberg’s AI in Finance Summit in New York on Oct. 15, Randich told the audience that FINRA, the self-regulatory organization that governs US brokers and broker-dealers, will continue to sustain a technology-neutral regulatory stance as AI becomes more embedded in firm operations: “The supervisory requirements and obligations of firms don’t change because of AI.”

Randich, who joined FINRA in 2013, acknowledged that certain “elements” are different. A decade prior, for instance, the organization discovered there were no solutions for model management and operations, requiring it to develop governance capabilities from “scratch.”

“This is a big area where FINRA is focused internally as well as externally, in looking at firms’ management of their models,” he said. “Do they have the right integrity, quality? Are they updated properly, and are the outputs accurate?”

FINRA is continuing to have conversations with firms to better understand how they are using AI and make sure there are no gaps in its guidance: “Thus far, [model ops are] really the only major change that we’ve seen. At the end of the day, firms’ obligations aren’t changing because of this technology.”

AI x FINRA. Randich told the crowd that FINRA has been using AI internally for the past decade, beginning when the organization transitioned to the public cloud.

“We started using open-source software for machine and deep learning, and we used it for market surveillance because that’s the area where we have a big data business issue, and we need very efficient, very smart surveillance,” Randich said.

The technology has presented a large opportunity for FINRA, which is often looking for anomalies and fraud within exabytes of data.

“We’re using it in dozens of areas within FINRA today. Some are [research and development] experimental, many are production,” Randich said. “But we’re using it everywhere, from surveillance to communications to inspecting advertising videos on firms’ websites to summarizing thousands and thousands of complaints that come into firms.”

The payback on these use cases have been fruitful. “The ROI is essentially becoming more effective and a better, deeper regulator,” Randich said. “It’s kind of like the police department, as an analogy. The police get more efficient. Doesn’t mean you have less police officers. You keep the same number of police officers to have less crime, and that’s what our goal is.”

Advice for CIOs. For CIOs at organizations that undergo a FINRA examination, Randich stressed the importance of having proper AI governance around models.

“Your AI and your business processes that are based on AI are only as good as the model and it’s very easy to let the model get away from you,” Randich said, adding that CIOs should maintain humans in the loop where appropriate and make sure there are employees who can be held accountable for AI processes.

“We want to make sure that there are people in the firms who are accountable and are looking at the results, the models, the processes that are moving to AI,” Randich said. “And as we move into agentic AI, it’s going to become even more important because things are going to become even more autonomous.”