While all that’s been going on, however, a tectonic shift in information technology has taken place featuring generative artificial intelligence systems, machine learning and rapidly evolving large language models that surpass the buzzy ChatGPT facility that is now so familiar to many. It’s now a sprint for these AI systems to develop superior capabilities to ingest information of all kinds, including images, and to create and manipulate databases, compile information in user-friendly formats for analysts and decision-makers, and deliver actionable analytics that are increasingly faster, cheaper and more insightful. Literally billions of dollars will be invested in this new AI technology in coming years.
The ownership of databases, analyses and related intellectual property scarfed up and refabricated by these systems is a burning issue that will spill into the governmental finance arena in short time. There is a non-trivial risk of concentrated monopoly or oligopoly control over powerfully AI-curated versions of what starts out as public information but quickly becomes private intellectual property when compiled, dissected, analyzed and commercialized by a proprietary machine learning system.
A Data Format Space Race
Of course, nobody has yet laid out specifically what financial information is actually deemed to be decision-useful in order to establish which data must be converted to a new format; that’s an unresolved first-order problem that leaves open the risk of unwarranted burdens on local governments during the initial implementation phase. All the oversight boards need to tread carefully.
What proponents of the FDTA had in mind when they lobbied Congress, standardization on the extensible financial reporting language platform that has become commonplace in the private sector, was only a first-stage rocket in this new space race. The federal legislation did not give XBRL a monopoly per se, specifying only the use of “structured” data formats.
Clearly, what most parties in the legislative process could never have anticipated last year was the possibility that existing financial reports using generally accepted accounting terminology may themselves already be computer-readable because of the new large language machine learning models that can read plain English typeset produced by word-processing software as well as alphanumeric images contained in the commonly used PDF documents that typically encapsulate governments’ audited annual financial reports. All of a sudden, “structured data” may ultimately prove to be little more than what we already have in place with conventional text documents that can be ingested by new AI systems with superior analytics already integrated with database utilities, without costly data entry hurdles.
That means there is now a realistic chance that all the laudable intent of the FDTA can be realized by many financial statement users without burdening those who prepare the documents. A regulatory case might develop that given the emerging AI technology, Securities and Exchange Commission (SEC) rules that prematurely compel what may soon thereafter become archaic structured data formatting are essentially a federally imposed waste of municipalities’ time and money. Sometime in 2024, somebody at the SEC may need to hit the pause button and rethink what’s going to be the most sensible regulatory path in light of the AI revolution, as the ground for assembling and analyzing data is shifting seismically right under their feet.
The Imminent Regulatory Challenge
Certainly these industry developments deserve a closer look than what lobbyists, congressional staff, the Securities and Exchange Commission and the Municipal Securities Rulemaking Board (MSRB) appear to have given them to date. It’s too soon to know whether the new GPT-4 and even-newer Claude2 AI models will scrape financial data (and images thereof) with sufficient accuracy and efficiency to equal or exceed the capabilities and efficiency of XBRL. But ultimately that will be far less important than the ability of evolving AI systems to ingest all this data — including decision-useful information that is not presented in basic annual financial statements, such as the footnotes and supplemental information, interim unaudited financial reports and budget documents — into a comprehensive database to then be sliced and diced in hundreds of different ways and packaged for myriad users as private goods, not public goods.
Thus, the regulators now need to formulate forward-looking terms of use for how data from reports they receive will be employed and converted commercially to prevent appropriation of public information by private parties. As in hockey, they need to skate to where the puck is going.
This new rival information technology may not be welcome news for companies that have laid down cash to produce conversion systems and hype the benefits of XBRL as the immediately obvious and only way to comply with the new federal law and anticipated regulatory requirements.
The next generation of multimodal AI systems should soon be capable of reading words and numbers from all kinds of plain language financial reports and PDFs both new and old, ingesting the key data, formatting it into databases, calculating customized financial ratios, organizing peer group comparisons and using historical data to formulate the predictive value of key statistics (just like football). AI could easily challenge the wisdom of buying into a modern-day Edsel of governmental fintech. Not that the Edsel was a bad car: If the switching costs of implementing XBRL now and migrating to a different technology later are minuscule, then it’s obviously no big deal and its fans can enjoy their day in the sun.
My concerns here could still be a false alarm, like a boy crying wolf. However, if myopic conversions to the less-flexible early-bird software “solutions” for FDTA are costly and time consuming, then a wait-and-see approach at both the regulatory and the local level may be wiser. This all reminds me of the early 1980s, when expensive “integrated financial management systems” operating on mainframe computers were procured by midsize municipalities, only to soon discover quite unhappily that microcomputers, cheaper software and ancillary data storage systems had swiftly leapfrogged that way of doing business and they had foolishly sunk capital into what quickly became “legacy” systems.
Profiteering from the Public’s Data
Putting aside the upcoming debate over how municipal numbers find their way into databases for analysis, the broader policy issue that public finance professionals and their membership associations should be focused on is who will own and control the next-gen databases and embedded analytical platforms that will inevitably be created from all this originally public information. It’s no longer merely about mandating easier access to selected numbers in individual financial statements for conventional decentralized analysis.
AI-empowered first movers could swiftly (by governmental standards, anyway) win a land grab in this corner of the world of intellectual property by building sophisticated machine learning systems that can ingest, store and — most importantly — analyze public-source financial information for use by bond investors, financial policymakers, oversight agencies and other users. Maybe the market leaders would do so benevolently as a public good, but what if not?
Even if the MSRB steps in and sets some industry ground rules for indelible watermarking and terms of use of its documents databases, my past public- and private-sector professional experience now leads me to believe that the time has come for municipal marketplace leaders to consider organizing a public benefit corporation to keep pace with this fast-moving industrial revolution.
The centerpiece of such a company would be a commitment by charter to establish the public finance community’s proprietary control of — and affordable, widespread access to — compiled financial information and analytics using the most advanced AI technology so that users could access state-of-the-art files only by license and subject to specific terms of use that representatives of the public interest can require. This is a second-step function that the various regulatory agencies cannot themselves fulfill.
Governmental providers of financial information — and their oversight agencies — should enjoy free use of the community-controlled databases for their own exclusive use and analytics. Municipalities could be exempted from reformatting their conventional annual financial reports themselves if they upload them to the collective database. Academic users would pay a small fee for the time-saving benefit of database compilation and embedded analytic tools and reports.
As for for-profit businesses, they would pay commercial fees for use as renters, not owners, of the database and its analytics under a licensing agreement with restrictive terms of use to prevent hijacking of these resources. They could download the derived output data and own the way they themselves slice and dice it, but they would never be able to claim ownership of anything but their own methods and the results that they themselves have uniquely created, with a royalty on their ensuing sales payable to the collective. New municipal market entrants such as women- and minority-owned businesses could be given pro-competition discounts. They can all buy a license to take fish from the lake, but should never own the water.
The contributors of financial data — not only governments, their agencies and professional associations but also any founding investors, platform licensees and the fee-paying users of financial information and analytics (such as bond rating and insurance companies and investment managers) — could equitably share the enterprise earnings, similar to the way a farmers’ grain warehouse cooperative operates for an agricultural community’s benefit.
For starters, public finance professional leaders should open a dialog with the SEC and especially the MSRB, as the latter should inherently be such a corporation’s co-founder, co-owner and member with a board seat (or the chair) because of the mutual interest relationships that would be necessary for this idea to gain traction. The amply funded MSRB could even become a cornerstone investor to help get this enterprise off the ground before FDTA-mandated reports start trickling in. MSRB’s Electronic Municipal Market Access website and databases would be core contributing facilities to get this protective apparatus established before it’s too late to shape the market.
The ultimate legal and business model is less important today than getting a discussion underway before it’s too late and high-tech market forces overtake the public interest.
Governing's opinion columns reflect the views of their authors and not necessarily those of Governing's editors or management. Nothing herein should be construed as investment advice.
Related Articles
