Securities Regulation Daily Wrap Up, SEC NEWS AND SPEECHES—IAC debates AI’s newfound influence in corporate disclosures, public markets, (Sep 10, 2026)

By Suzanne Cosgrove

AI may ease analytical burdens for filers but it does not give the Commission latitude to ignore its principle of materiality, SEC Chaiman Paul Atkins says.

The SEC’s Investor Advisory Committee on Thursday examined how AI-driven technologies are reshaping public company disclosures, with the panelists focusing not on whether the technology has changed the process, but on the extent of that change and what, if any, rulemaking might be needed in response.

In prepared remarks delivered to the IAC panel, SEC Chairman Paul Atkins said that while AI “holds considerable potential, it should serve as a complement to—not a substitute for—human judgment. Its susceptibility to errors and hallucinations remains a significant concern in the context of disclosures on which investors rely to make informed decisions.”

Looking for context. “Are AI-enabled technologies narrowing or widening the relative gap between large and small entities’ experiences and expenses associated with disclosure and structured data?” asked SEC Commissioner Hester Peirce in her remarks to the group. “If the divide is widening, how can we respond?”

Regulating prematurely can deprive Americans of the benefits of technological innovation, Peirce noted. “That said, are there any issues presented by AI-enabled technologies that are currently ripe for Commission review?”

“From an investor perspective, retail investors are disadvantaged [compared with] institutions who can deploy AI at scale,” said panelist Campbell Pryde, president and CEO of XBRL US Inc. XBRL permits automated sharing of financial data using tags to sort data use.

Institutions can get better rates, “so they're always going to be going to disadvantage retail investors unless there is a lot of competition and the cost of AI comes down,” Pryde said. “And then, inevitably, we don't know where these data processing costs are going to come out given the amount of money that people are spending on AI at the moment.”

Data quality. In her presentation to the committee, Julie Marlowe, assistant director, Office of Data Standards and Innovation, SEC Division of Economic and Risk Analysis, spoke to the speed at which data analysis is available with the new technologies, with some processes taking just seconds when using standard tagging.

At the same time, Marlowe said that although her office is not allowed to contact individual filers, it is working on data quality generally. “When we see a systematic issue, we issue staff observations and guidance and data quality reminders for the public to see. You'll often start with: Here's the data quality that should be noted, here's how you can fix it, here's the guidance.

“We've seen scaling errors for public float. So, for example, a small company with trillion in public float, which does not make sense,” she noted. “We've also seen some filers missing three zeros in the XBRL data, and when we see those issues, you may wonder what does the finance reporting process look like [at that company]? Is it including checking the XBRL data?”

Is speed an edge? “AI is not the great democratizer of investing,” said Gregory Landegger, principal at Parsons & Whitmore. “What is unquestionably delivered is speed. Research that took a weekend of reading 10ks now takes seconds. Financial models can be generated from a single prompt. Research transcripts and data rooms can be processed at a scale no human team can match,” he said.

“But speed is no longer the edge it once was,” he said. “When everyone processes the same information faster, the differentiator moves elsewhere. “Watching how family offices and investors use these tools, I've come to a simple division. Investors' use of AI falls into two separate categories: deliberative tasks and evidentiary tasks.”

“Deliberative tasks are where AI truly shines,” Landegger said. “These are tasks where the output is a question, a challenge, or a perspective rather than a fact. Used this way, AI expands your field of reasoning by asking questions you may not have thought of yourself.

“Evidentiary tasks are different,” he said. “Here, the output must be correct, traceable to a source, and identical every time the question is asked. I'm talking about repeatability. Predictive systems do not ensure this. When I ask an AI to analyze an earnings report, I need the answer to be the same every time. … Predictive systems do not ensure this,” he said.

“Many suggest that large language models will improve, and the problem will fade. And I believe accuracy will improve, [but] repeatability will not, because these systems … are probabilistic by design. A better model gives you a better distribution of answers, but it is still a distribution. For a deliberative task, this is fine. For an evidentiary task, it falls short.

AI and markets. Given this limitation, the rise of AI trading deserves our attention, Landegger said. “Retail brokers are beginning to let customers delegate trades to autonomous agents with set budgets and risk limits. Consider what this means: the decision to buy or sell is being made by a probabilistic system acting on its own readings of filings and news with no record of what it reads or how it acted.”

Looking at the prediction markets, AI is even further along, he said. “According to some reports, AI already operates a third of active accounts and a majority of the most profitable ones. Humans still may be making the right bets, but if AI executes faster, humans are going to lose.”

RegulatoryActivity: AINews ExchangesMarketRegulation FiduciaryDuties FormsFilings SECNewsSpeeches