AI models are helping investors speed up analysis of lengthy corporate filings while savvy companies are exploiting the weaknesses of the new tools.
Investors are using artificial intelligence models to speed up research, easing the laborious process of comparing documents across years to find changes in information. Some companies, meanwhile, are drafting their filings in a way that AI models would be more likely to miss or downplay information that executives don’t want to highlight.
The Securities and Exchange Commission under the Trump administration has contended that required corporate filings such as annual financial reports can be far too long with details that are insignificant to investors. Chairman Paul Atkins is mulling curtailing executive compensation and risk factor disclosures.
Investors are using tools like
Information previously could be buried deep in a corporate document, making it opaque, said Wesley Gray, CEO of Alpha Architect, an asset management firm. “But now everyone has a supercomputer.”
Regardless of how AI is used by investors, SEC officials say shortened disclosures can serve a broader purpose: giving companies incentives to go public and stay in the market amid a decades-long slump in the number of public businesses.
The SEC is trying to ensure disclosures “are useful and not overwhelming,” said Joshua White, chief economist and director of the agency’s Division of Economic and Risk Analysis, in a statement.
“One thing we have to keep in mind is disclosures are not costless to produce for companies,” White said. “It not only takes the time to produce it, you have to verify it, you have to stand behind what it says.”
Faster Tools
For decades, investors had to “control-F” their way through a corporate filing to find information, said Corey Hammill, senior vice president of financial data at AlphaSense, a market intelligence platform.
Investors using AlphaSense’s AI platform can find out, for example, if
AlphaSense uses AI to extract details from SEC filings and earnings call transcripts, among other documents. Its tools are used by companies including
“In my experience as somebody that was in capital markets for most of my career, time is my most precious resource,” Hammill said.
The point of reading an annual report is to make an investment decision, said Doug Clinton, a co-founder and managing partner at Deepwater Asset Management.
When it comes to these decisions, some AI models broadly “are already capable of being better than the average human professional manager” Clinton said.
Still, there are risks of hallucinations and errors.
Corporate Impacts
AI models also can be tricked.
There are ways to write a company disclosure to get favorable responses from LLMs, said Bradford Levy, a University of Chicago professor who researches the intersection between AI, information processing, and financial markets.
“I have a class where I teach people how to use AI to process financial information, and this is literally a lab that we have,” Levy said. “So, the students aren’t allowed to rewrite the content of the disclosure, but they can shuffle paragraphs around so they can move section seven to the front of the 10-K, things like that.”
The result: Some AI models are less likely to look at the content in the middle of a document.
Those parsing company documents have to anticipate that firms are going to behave a certain way and figure out how to write prompts and interact with disclosures through the lens of AI to get more robust answers.
In Levy’s class, students in another section of the lab have to try to unwind the positive bias that other students attempt to put in the disclosure so that companies cannot game the situation.
Moving forward, companies might find AI tools useful in their own work.
Hammill said companies use AlphaSense’s tools to stay on top of their industries and competitors, as well as monitor what the market is saying about them.
“It’s a full-time job—hence why they have these departments—to keep up on what the market is saying about you and what they’re saying about your competitors,” Hammill said.
SEC and Long Documents
AI’s new role makes the push against voluminous company documents increasingly irrelevant, Levy said.
“The SEC would be well suited to think more about the interaction between how investors looking at disclosures through AI, the lens of AI, what are the strategic interactions there of managers trying to bias the statements,” he said.
The burden of lengthy disclosures that don’t provide useful details to investors can discourage companies from going public, the SEC has argued.
“At the end of the day, even if investors have all these tools available to them to process data, if the public companies are not there, there’s not going to be any data to process,” Sebastian Gomez Abero, a deputy director in the SEC’s Division of Corporation Finance, said in a statement.
Although humans still need to be involved because of the dangers of hallucinations, AI can look at more data based on trends and across sectors and companies, said Andrew Behar, CEO of As You Sow, a shareholder advocacy group.
“It strikes me that the SEC is trying to suppress information at this moment where we can finally handle a lot more information,” Behar said.
Bloomberg Law is a legal technology provider.