The IRS is going after fewer of the largest, most complex businesses in the country—and when it does try, it often runs out of time, staff, or both. That’s the conclusion of a new audit from the Treasury Inspector General for Tax Administration, the independent agency that examines how the IRS is handling audits of large partnerships—entities that can generate enormous income but are notoriously difficult to police.

Over the past decade the number of large partnerships—those with at least $10 million in assets—more than doubled, but during the same period the rate at which the IRS examines them has collapsed, falling from 2.7 percent to less than 0.1 percent. In other words, as the population of these complex entities has surged, the chances they’ll be audited have all but disappeared.

Partnerships don’t pay taxes directly. Their income flows through to partners, often through layers of entities that can obscure where money is made and where taxes are owed. The IRS already estimates that at least $42 billion in income from these types of entities goes unreported each year—and acknowledges that figure is likely understated.

But when the IRS does manage to dig into these returns, the payoff can be substantial. One study cited in the report found that for every dollar spent auditing complex partnerships, the government recovers about $20. The problem isn’t whether enforcement works. It’s that it’s barely happening.

In one initiative, the IRS identified 483 large partnerships with clear discrepancies in their balance sheets and sent them letters asking for an explanation. Of those, 163 never responded, and of the responses received, 182—about 57 percent—were rejected as incomplete or insufficient. Under normal circumstances, that would trigger audits. But that didn’t happen.

By April 2024, the IRS decided it would not pursue examinations on any of these cases—despite nonresponses and rejected explanations—because it lacked the time and resources to act before the statute of limitations expired. The agency walked away not just from the nonresponsive partnerships, but from all of them. In doing so, it also created a basic fairness problem: some partnerships spent time and money responding to the IRS, while others ignored the letters entirely, and none were examined.

In a separate effort, the IRS turned to artificial intelligence to identify the highest-risk returns. The model analyzed more than 282,000 large partnership returns for Tax Year 2021 and flagged 1,617 for further consideration. From that group, the IRS selected 150 for deeper classification review, ultimately choosing 78 for examination — a number that grew to 82 after examiners identified related transactions.

As of December 31, 2025, 43 of those exams were still ongoing. Of the 36 that had been closed, 33 — about 92 percent — resulted in no changes to the return.

But the model was run only once. As a result, 2,204 additional qualifying returns were excluded — not because they posed low risk, but because they were filed after the model’s cutoff date. The IRS attributed running the model only once to limited staff and time, even though it typically runs selection models multiple times in other programs to ensure broader coverage.

This same problem shows up everywhere in the report. The unit responsible for these audits lost more than 20 percent of its workforce in 2025 alone, including a large share of the revenue agents who actually conduct examinations. Even before those losses, the IRS had already concluded it lacked the resources to keep up with the growth of large partnerships.

The result is a system that can identify problems but often can’t act on them. Returns with obvious discrepancies go unexamined. High-risk cases fall outside the selection window. Entire initiatives are abandoned midstream because the clock runs out. The IRS is not choosing to ignore these cases—it simply doesn’t have the capacity to pursue them.

The House Appropriations Committee is now moving forward with a Financial Services spending bill that would cut IRS funding to $10.2 billion next fiscal year — a roughly 9 percent reduction from the current $11.2 billion. Efforts to increase enforcement funding were rejected, and the proposal instead cuts enforcement by about $1.4 billion, bringing it from nearly $5 billion down to $3.6 billion — a reduction of more than a quarter. Notably, the House bill goes further than even the Trump administration requested: the administration’s own budget proposed cutting enforcement to $4.1 billion, an 18 percent reduction, while the House bill cuts deeper still.

Supporters of the cuts argue the agency should do more with less. But the TIGTA report shows what “less” already looks like in practice. It looks like AI getting enforcement into the red zone and then fumbling the tax ball. It looks like hundreds of flagged returns never being examined. It looks like audit selection tools that can’t be fully used because there aren’t enough staff to follow through. It looks like the IRS identifying risk and then walking away from it—not by choice, but because it runs out of time.