MSRB adding AI to EMMA

The Municipal Securities Rulemaking Board is currently wrestling with adding AI to its EMMA system in preparation for a machine readable future for financial disclosure.   

“Last year, we received almost 12,000 primary market submissions from you, mostly in the form of official statements,” MSRB CEO Mark Kim told the Government Finance Officers Association’s Debt Committee. “It may surprise you that today, in 2024, we manually touch every single one of your OS’s.” 

The comments came during a Monday meeting of the committee that launched as part of GFOA’s winter meeting in Washington. The MSRB revealed how artificial intelligence is being leveraged to upgrade the EMMA system primarily within the G-32 document. Nearly half of the board’s $45 million budget goes towards technology and data efforts.   

“Last year, we received almost 12,000 primary market submissions from all of you mostly in the form of official statements,” said MSRB CEO Mark Kim. And it may surprise you that today, in 2024, we manually touch every single one of your OS’s.” 

Dan Nelken/Dan Nelken

The building blocks of the AI methods being employed harken back the early days of data entry. “This is not generative modeling or generative chat GPT,” said Brian Anthony, chief product officer, MSRB. “This uses a concept called supervised learning, which means that we have to go in and teach the model what it is that we’re looking for.” 

Problem areas for the process include standardizing how dates are recorded and the organization of tables of numbers. 

“The good news is that underwriters, generally speaking, do an excellent job,” said Kim. “Underwriters are required to provide this information to us, if they fail to do so, or if they fail to correct data that we identify as potentially wrong, we provide that information to the enforcement agencies.” 

The new experimentations are also designed to boost publicly available information on obligors that is currently restricted in EMMA. 

The evolution of EMMA and the quality of the data being entered also relates to how it will intersect with the standards being developed by the Securities and Exchange Commission imposed by the Financial Data Transparency Act. The law, currently in a phase in-process, will require issuers to produce financial disclosures in a machine-readable format.

“As we look forward to the SEC and Treasury rulemaking on FDTA, establishing data standards, it’ll be really interesting to see whether those types of (accuracy) issues are addressed or not,” Kim said. “We are providing technical assistance to Treasury and SEC. Structured data, semi structured data, unstructured data, we receive it in a wide variety of kinds of formats.”

The FDTA represents a point of pain and opportunity for the GFOA as the law forces issuers to convert some disclosure documents into a yet-to-be-identified machine-readable format. The GFOA does have a plan in place to take advantage of the upcoming public comment period starting in June to help finesse the fine details of the law. 

“You can bet your britches, we are going to make sure that there are comments coming in from every GFOA member so that they understand the data standards, and what type of data standards will be used and what type of data standards will be deployed,” said Emily Brock director of the Federal Liaison Center, GFOA. 

Brock has also been keeping her eye on the major tax package currently winding its way through Congress. “There was a lot of stuff in that tax package that was related to the 2017 tax act,” she said.  Some of the features that they have put inside of that tax bill are relative to corporate transactions and corporate taxation, not necessarily distinctly our issues, but there’s one distinctly our issue that found its way which is Low Income Housing Tax Credits.