Paul Jones & Emily Hollingsworth, Unwatched: How State Audit Selection Systems Bypass Oversight, Tax Notes (Apr. 14, 2026) (quoting Jacob Goldin (Chicago) and Bob Probasco (Texas A&M)):

One obstacle to proper oversight is a lack of clarity about what qualifies as AI — and with it, a lack of awareness about the need to review automated systems as attention fixates on more advanced AI, like deep-learning-powered large language models.

In 2025 Tax Notes surveyed 49 state and jurisdictional tax agencies to ask whether the agency in question used machine learning or generative AI in its tax operations, including to identify tax returns for audit. Of the 36 agencies that responded, 22 said they didn’t use AI at all in their tax operations.

Seven agencies said that they did use AI programs, ranging from chatbots to machine learning models. Six agencies indicated that they were interested in looking into the use of AI or machine learning systems, including some of the agencies that said they currently didn’t use AI. Three agencies didn’t disclose whether they used AI.

Although many states answered that they weren’t using AI systems, Ryan Minnick of the Federation of Tax Administrators said most state tax agencies already use legacy AI tools — such as machine learning and advanced data analytics — for tax return processing and fraud detection and prevention. . . .

“There’s internal confusion about what is and is not AI,” Nonnecke said.