Opening Remarks at the Roundtable on Machine-Understandable Data in the Age of Super Intelligence

Good morning, and welcome to today's roundtable on machine-understandable data and super intelligence. Thank you to everyone who is here in person and to those who are joining online. I regret that I cannot be with you in person, but I appreciate the opportunity to share a few thoughts.

The Commission's disclosure-based regime rests on the premise that investors and markets make better decisions when companies provide timely, accurate, honest and financially material information. For much of our history, the reader of disclosures was a person with a printed document. Then it was an analyst with a screen. Today, increasingly, the first reader of a filing is a machine. Whether it is a quantitative model, a large language model, or other super-intelligence tool, these machines extract, ingest, and interpret disclosures at a speed and scale that humans cannot match. As super-intelligence tools become more widely used, that trend will only accelerate.

So, how can machine-readable data complement super intelligence? Does a format that was introduced over two decades ago still have value in an SI environment?

The Commission mandated XBRL starting in 2009,[1] and we have since required Inline XBRL[2] and expanded structured data requirements to additional filings.[3] While these requirements carry compliance costs that we as regulators should be mindful of, they have also led to significant benefits. Structured data is critical to attracting research analyst coverage, especially for smaller companies, even when that coverage is automated or aggregated in large databases. In other words, machine-readable data can increase visibility, which in turn lowers the cost of capital by making information easier to find, compare, and analyze for investors.

Super intelligence, far from making XBRL irrelevant, may make this connection even stronger as machine-readable data can feed the SI tools with more reliable and consistent data inputs. SI tools can potentially analyze smaller companies that were historically overlooked or spot patterns and trends in data in ways that prior technology did not. The technology is advancing so quickly that none of us can say what these tools will be capable of in a few years.

Today's discussion is a valuable step in thinking about how SEC filing data can be better utilized in the age of super intelligence, including whether there are new methods that are superior to XBRL. The first panel will explore how market participants are currently using these tools. The second panel will examine the risks and costs the Commission should consider and how we might better facilitate progress. I encourage everyone to be candid, including about where the Commission can do better. Facilitating capital formation is part of our tripartite mission, and high-quality structured data can help advance it.

I want to thank Dr. Joshua White, Julie Marlowe, Valerie Szczepanik, and the staff of the Division of Economic Risk Analysis who organized this event, as well as our panelists for sharing their time and expertise. Thank you.