Artificial intelligence has the potential to revolutionize the disclosure process, however, the proposed benefits might not outweigh the accompanying risks, according to analysts.
AI has tremendous utility for issuers and counsel, panelists agreed during The Bond Buyer Tech Forum Thursday. They praised the tool's ability to research and collect vast amounts of information while packaging that data in a digestible format for users.
"There's so much information out there about us, and there's so many different varieties of ways that people look at risks," said Emily Brock, director of the Government Finance Officers Association's Federal Liaison Center, said during a panel. "For us to have a tool that's like a large-mouth bass, sucking up all of those pebbles at the bottom of the river and processing them … we have a way to understand all of that data."
From a disclosure counsel perspective, AI does not just serve as a research tool. It can provide "a second set of eyes," catching mistakes in document drafts and proofreading to ensure clarity, said Noah Lebowitz, counsel at Nixon Peabody.
"It can have fantastic purposes in operational roles in assisting people who do prepare disclosure so that they can spend more time thinking about the actual content of the disclosure than proofreading," he said.
The panelists commended AI's ability to maximize time and increase efficiency with low-stakes tasks, however, they emphasized the risks associated with relying too much on the tool and allowing it to encroach upon aspects of disclosure.
The strongest criticism the panel levied related to creation of original content, specifically drafting disclosure documents.
There is a reason why AI has "very little to no role in drafting those documents and creating those documents today," Lebowitz said. The tool is too likely to generate information that is inconsistent with the needs and wants of the issuer, he added. Right now, there is too great a gap between credible documentation creation and AI's current research capabilities.
"You must verify everything that the AI does, that includes holding a pen on your document," Lebowitz said. "The risk-reward is: does AI save me more time in preparing that document than it otherwise would have, knowing that I now need to go in and verify all of the work that the AI has done?"
Nikolai Sklaroff, capital finance director for the San Francisco Public Utilities Commission, spoke about the danger of widening the universe of information. AI can scrub the internet for information that is unavailable to an issuer.
If an investor uses AI to base its perception on AI-collected research, there could be a scenario where the investor believes an issuer was hiding information because the AI collected from more sources, he said. Moreover, if an issuer uses AI to draft documents and that model pulls from sources of information outside of an issuer's known universe, then unverified information could confuse and intermingle with verified information — which presents problems for the overall disclosure process, Sklaroff said.
"Seeing investors accessing data sets that we, as issuers, don't have access to ... I worry about this divergence of what we have access to and what we are able to disclose," Sklaroff said.
The panel acknowledged AI's implementation in the industry is a reality, but it should be primarily used to manage and navigate non-AI-generated material.
Sklaroff emphasized the importance of understanding how new technologies interact with the marketplace and shared four areas of consideration that he deals with when considering AI implementation: a firm's specific AI needs, relationships that exist with AI companies that are marketing new products, understanding who you are working with, and how certain AI programs aggregate information.
Brock referenced a Bond Buyer
Brock took it a step further. She said, using AI during the disclosure process is like trusting the tool to tell an issuer's story. Even though "AI is great at collecting information … the issuer needs to control their stories."
