AI vs. SEC

Writing the boilerplate verbiage that bolsters preliminary official statements used in bond sales is now being filtered through AI strainers and attracting attention from the Securities and Exchange Commission.

"It's easy to go down the rabbit hole of saying, 'how can we change this language, how can we use this language,' and that can very much lead you astray if you let that be your only consideration," said Dave Sanchez, director of the SEC's Office of Municipal Securities.

Sanchez explained that the antifraud provisions of the federal securities laws dictate that any disclosure be accurate, so issuers shouldn't get caught up in using AI to make their securities sound more attractive to investors if that language risks becoming misleading. "SEC is greater than AI," Sanchez said.

The comments do not reflect official SEC policy and came during the Government Finance Officers Association meeting of the debt committee on Saturday in Chicago.

Sanchez recommended the primary consideration in disclosures are accurately describing the credits.

The influence of AI is leaking in from all sides of the bond sale and influencing anybody reading the statements.

"AI is being used by every large asset manager," said Colin MacNaught, CEO & Co-Founder, BondLink.

"It's not just the credit analyst, it's the traders, it's the portfolio managers, automating their workflows. That relies on easy and reliable access to data and issuer financial reports. That's just very important for issuers to know as you think about your disclosure program going forward."

MacNaught advises against using different words to describe the same concept and steering around anything that sounds ambiguous.

"Humans can read between the lines," he said. "AI is just reading the line. If a portion of disclosure is trying to be prudent and careful about how you're describing a certain portion of your credit story, that could be misinterpreted by the AI bot as a risk."

Ratings agencies are rolling AI into their operations and research functions. Academics are building AI-powered models to replicate the ratings process.

Per a recent paper published on the Social Science Research Network, "Traditional representations of firms use accounting and financial market data, but investors use richer information sets. Theoretically, portfolio holdings contain all relevant information for asset prices, recoverable under empirically realistic conditions."

Although AI can contribute to risk factors in disclosure it can also help issuers tell better stories.

"Those issuers who understand what investors need now, could see more interest from the buy side," said MacNaught.

"Centralize and organize your financial reports and data ? that essentially tells AI where to search. Be your own source of truth and minimize the risk that AI will pull from unreliable sources of information."

MacNaught believes leveraging machine learning tools to get the message right could help smaller issuers level the playing field.

The tide of AI is sweeping across all the financial markets while the final results remain unknown.

"Seven tech sites that are claimed to be AI driven now account for something like 38% of the S&P 500," said Justin Marlowe,professor, University of Chicago.

"A study that the Dallas Fed came out with a few months ago looked at the long-term implications of AI from the economy, and their conclusion was the title of the slide, "AI might either destroy humanity and scarcity or lose GDP by 21 basis points.' In other words, no damn idea."