The IRS Flagged 483 Big Partnerships. It Audited None.
Summary
A March 2026 audit by the Treasury Inspector General for Tax Administration found the IRS sent balance-sheet-discrepancy letters to 483 of the country's largest partnerships in 2023, then decided in April 2024 to examine none of them, citing limited staff and a ticking three-year deadline to act. It's one symptom of a wider collapse: the IRS's audit rate for partnerships with at least $10 million in assets fell from 2.7% in 2011 to under 0.1% in 2023, even as the number of those partnerships grew 138%. The IRS had banked on $79.4 billion in Inflation Reduction Act funding to raise that rate ten-fold by 2026; Congress has since rescinded $53.5 billion of it across four laws, leaving about $26 billion, and the enforcement unit lost more than a fifth of its staff in 2025.
483 warnings, zero audits
In October 2023, the sent Letter 6585⧉ -- a "soft letter," voluntary compliance correspondence that flags a problem without opening a formal examination -- to 483 large partnerships whose Form 1065 balance sheets didn't reconcile: the assets on the books didn't match liabilities plus partners' capital, a red flag for misstated income. Partnerships had 30 days to fix the discrepancy, explain it, or ask for more time. Of the 483, 163 never responded at all; of the 320 that did, the rejected 182 responses as inadequately documented and accepted 138. By the 's own account, that left 345 of the 483 -- the silent and the rejected -- as unresolved, exam-worthy cases.
In April 2024, the 's Large Business and International Division decided to examine none of them. The reason wasn't insufficient evidence -- it was the clock: examiners generally can't open a case with fewer than 12 months left on the three-year statute of limitations to assess additional tax, and the duplicative, sequential internal review⧉ the built to process the 483 responses ate the time that would have been needed to act on them. called this a fairness problem in its own right: partnerships that spent time and money responding to a federal letter got no resolution, while those that ignored it faced the same non-consequence.
View data as table
| No response | 163 | |
|---|---|---|
| Response rejected as inadequate | 182 | |
| Response accepted | 138 | |
| Actually examined | 0 | In April 2024 the IRS decided not to examine any of the 483, citing limited resources and the assessment statute of limitations |
The AI-driven audit push is finding mostly nothing too
The 's other large-partnership initiative, the Large Partnership Compliance (LPC) Program⧉, uses an AI risk-scoring model to flag the biggest, most complex returns for examiners -- the kind of case that can involve hundreds of thousands of partners across multiple tiers. Run against 282,884 Tax Year 2021 returns, the model surfaced 1,617 candidates; the ultimately selected 82 for examination, about 0.03% of the returns it modeled. But because the model was run only once, 2,204 qualifying returns filed after February 2023 -- 0.8% of the population -- were automatically excluded from consideration, with no second pass to catch them.
As of December 2025, 43 of the 82 LPC exams were still open, 3 hadn't been assigned to an examiner, and of the 36 that had closed, 33 -- 92% -- ended with no change⧉ to the tax owed at all -- results itself calls preliminary, since most of the 82 exams are still in progress. Still, that 92% is worse than the partnership population overall: the 's own no-change rate across all partnership exams, not just the LPC Program's, ran 44% in Tax Year 2014 and 47% in Tax Year 2021 -- meaning the flagship, AI-targeted effort is missing more often, on the cases it has finished, than the untargeted baseline it was built to beat.
That doesn't mean the effort is worthless -- a study cites, in which the itself participated, found that average adjustments in complex partnership exams run eight times higher than in large corporate exams, and that the collects $20 in revenue for every dollar it spends⧉ examining them. doesn't itself draw a conclusion from that ratio, but it implies a real return being left on the table when exams don't happen at all -- and it's also the standard the itself set: in its 2023 strategic plan for the enforcement money, the agency projected it could raise the partnership examination rate ten-fold, from 0.1% in Tax Year 2019 to 1.0% by Tax Year 2026.
View data as table
| Tax Year 2011 | 2.7% | 140,577 large-partnership returns filed that year |
|---|---|---|
| Tax Year 2023 | 0.1% | 334,686 returns filed (+138%); the IRS states the true rate is 'less than' this value |
The money for that goal got cut, and so did the staff
That 10x target relied on money the no longer has. The 2022 Inflation Reduction Act appropriated $79.4 billion⧉ in supplemental funding, a portion earmarked for exactly this kind of enforcement. Four subsequent laws -- the Fiscal Responsibility Act of 2023, the Further Consolidated Appropriations Act of 2024, a 2025 continuing-resolution rescission, and the Consolidated Appropriations Act of 2026 -- clawed back a combined $53.5 billion, leaving roughly $26 billion. separately states that $41.8 billion of that was cut specifically "from the enforcement funding activity" -- a narrower figure describing the enforcement-specific slice of IRA funding rather than the full $79.4-billion-to-$26-billion drop; the report doesn't say what that enforcement line item's original size was.
Separately, the Pass-Through Entities program -- the unit that runs both the soft-letter campaign and the LPC Program -- had 1,079 employees in January 2025 and lost more than a fifth of its staff⧉, nearly half of them revenue agents, by December 2025 to the Deferred Resignation Program, retirements, and terminations. The told it has not estimated what the funding cut and the staffing loss do to the 10x goal, only that LB&I management "will need to reassess their exam coverage goals for large partnerships following the reduced staffing" -- no date attached.
View data as table
| Original appropriation, Aug. 2022 | 79.4 | |
|---|---|---|
| After Fiscal Responsibility Act 2023 (-$1.4B) | 78 | |
| After Further Consolidated Appropriations Act 2024 (-$20.2B) | 57.8 | |
| After Full-Year CR 2025 (-$20.2B) | 37.6 | |
| After Consolidated Appropriations Act 2026 (-$11.7B) | 25.9 | IRS reports this as 'approximately $26 billion' available as of February 2026 |
What the IRS agreed to fix
made two recommendations, and the agreed to both: stop letting duplicative review steps eat the statute-of-limitations clock on future compliance initiatives, and run the LPC Program's selection model more than once a year so partnerships that file later aren't automatically waved through. The IRS committed⧉ to developing procedures for the second fix but set no implementation date -- the report says only that the agency "anticipates increasing the frequency of the LPC modeling" as the program matures. Neither recommendation addresses the deeper problem documented but didn't have the mandate to fix: the funding and staffing behind the whole enforcement push are smaller than when the set its own targets, and nobody has recalculated what those targets should be now.
- The 483 audits didn't happen because of a deadline the built for itself, not a lack of evidence. found the balance-sheet discrepancies were real and flagged; what killed the follow-through was a duplicative internal review process that ran out the statute-of-limitations clock before anyone could act on it.
- Even the AI-targeted flagship effort mostly finds nothing -- which cuts two ways. 92% of closed Large Partnership Compliance exams ended with no change, suggesting either the model needs sharper targeting or that complex partnerships are harder to catch than the case-selection process assumes. Either way, it's evidence against the idea that simply pointing more resources at the current process would close the gap on its own.
- The set its own bar for what enforcement should look like, then lost the money and staff to clear it. The agency's own 2023 plan projected a ten-fold examination-rate increase by 2026. Since then, Congress has rescinded $53.5 billion of the $79.4 billion behind that plan -- separately cites $41.8 billion cut specifically from the enforcement funding line -- and the enforcement unit lost more than 20% of its staff. The has not said what its revised target is.
The examination-rate figures (2.7% in TY2011, under 0.1% in TY2023) and the filing counts come from -provided data as reported by ; 's report presents them as a chart rather than a full year-by-year table, so this piece uses the two endpoint years states directly rather than interpolating intermediate years. "No-change" examinations are audits that close without any adjustment to the tax owed -- they consume examiner time and taxpayer time alike but produce no revenue, which is why and the both treat a high no-change rate as a case-selection problem worth fixing rather than a neutral outcome. The $79.4 billion original IRA figure and the four rescission amounts are stated in the report's own footnote documenting the underlying public laws; this piece independently re-sums them and confirms the arithmetic matches the 's rounded "approximately $26 billion" figure.
Sources(1) ▾
- Treasury Inspector General for Tax Administration (TIGTA), The IRS Has Yet to Develop a Successful Strategy for Examining Large Partnership Returns (Report 2026-308-011) (2026-03-18) — 's final audit report assessing whether the is using available information to identify large-partnership noncompliance, covering the Letter 6585 soft-letter balance-sheet campaign and the AI-driven Large Partnership Compliance (LPC) Program. Includes the 's own management response. Fetched directly from tigta.gov and converted with pdftotext -layout; no Wayback snapshot was indexed as of capture time (checked via the Wayback availability API), so the capture link points directly at the publisher's PDF. tigta.gov · original document
Comments
Always open. Logged-in readers can annotate paragraphs in place.
The Treasury Inspector General for Tax Administration⧉ -- , the 's own independent watchdog -- found that the number of partnership returns filed by partnerships holding at least $10 million in assets grew from 140,577 in Tax Year 2011 to 334,686 in Tax Year 2023, a 138% increase. Over the same period, the 's rate of examining those returns collapsed from 2.7% to less than 0.1% -- a decline of more than 27-fold. The stakes aren't abstract: the 's own Tax Gap estimate puts unreported income from partnerships and similar pass-through entities at $42 billion a year⧉, a figure the agency itself says likely understates the true total because sophisticated partnership noncompliance is hard to detect without examining the returns the has stopped examining.