Treasury rejected two-thirds of GAO's tax-rulemaking fixes
Summary
A GAO review of tax rulemaking found Treasury and IRS proposed 236 regulations and finalized 231 between January 2017 and March 2026, with 28 crossing the $100 million economic-effects threshold GAO calls economically significant. For those 28, the report says Treasury and IRS consistently analyzed alternative regulatory designs but inconsistently produced the specific cost, benefit, and revenue estimates leading practice calls for. GAO issued six recommendations to fix that gap and a separate strain — rulemakings drawing tens of thousands of public comments, some AI-generated, that IRS has no policy to sort — and Treasury agreed with two of the six, rejecting four.
Nine years, one soft analytical spot
The 236 proposed and 231 finalized⧉ regulations span the tax-law overhauls of 2017, 2022, and 2025. 's focus falls on the narrower set that matters most financially: the 28 rules crossing the $100 million economic-effects line that triggers the highest tier of regulatory scrutiny. That works out to roughly 1 in 8⧉ of everything Treasury and proposed to change in the tax code over the period — the share of tax rulemaking carrying the biggest dollar stakes.
View data as table
| Proposed | 236 | regulations proposed over the 9-year window |
|---|---|---|
| Finalized | 231 | regulations finalized over the same window |
| Economically significant | 28 | carrying $100 million or more in economic effects — the subset GAO's analysis-practice findings target |
What GAO says gets skipped
Federal guidance from the Office of Management and Budget⧉ recommends several practices for analyzing a regulation before it's finalized: comparing alternative designs, and quantifying the specific costs, benefits, and revenue each alternative would produce. found Treasury and reliably did the first — and that doing it paid off, making it easier for taxpayers to claim new tax benefits in the cases reviewed. The second habit — specific cost, benefit, and revenue estimates — was applied inconsistently, which says leaves Treasury and making regulatory-alternative decisions without the numbers that would make those decisions defensible.
A comment system with no policy for its own scale
The same report found Treasury and "are not following leading practices for public engagement in rulemaking⧉ and risk not being prepared to address voluminous public comments, sometimes tens of thousands, on proposed tax regulations." is contending with a newer problem on top of the volume: AI tools that can generate slightly different versions of the same comment for each submitter, which says makes it "more difficult for to identify duplicate comments" — and has not developed a policy for identifying, documenting, or otherwise handling mass or AI-generated comments. The public-hearing side of engagement has its own gap: hearings are held only in Washington, D.C., and offers dial-in audio only, not video conferencing, for remote participation.
Six recommendations, four rejections
's six recommendations split along the two problems it found: three direct Treasury to add specific cost, benefit, and revenue estimates to its regulatory economic analyses; three direct to build policies for mass/AI-generated comments and to modernize hearing participation and documentation. Treasury agreed with two of the six and disagreed with four⧉; 's response is on the record: it "maintains that implementation of all six recommendations would improve the development of tax regulations." As of the report's July 21, 2026 release, every one of the six was still listed Open, with waiting to confirm what action, if any, either agency takes.
View data as table
| Agreed | 2 | Treasury agreed with 2 of the 6 recommendations |
|---|---|---|
| Disagreed | 4 | Treasury disagreed with 4 of the 6; GAO maintains all six should be implemented |
- 28 tax regulations crossed 's $100 million economically-significant threshold between 2017 and 2026, out of 236 proposed — the highest-stakes slice of nine years of tax rulemaking.
- Treasury and reliably compared alternative regulatory designs for those 28 rules, but inconsistently produced the specific cost, benefit, and revenue estimates says should inform the choice between alternatives.
- Public comment on proposed tax regulations can run into the tens of thousands, increasingly including AI-generated submissions, and has no policy for identifying or handling either.
- issued six recommendations to close both gaps; Treasury rejected four of them, and all six remain open as of publication.
's economic-analysis findings rest on four selected regulatory economic analyses it compared against 's recommended practices, chosen because they were the most relevant examples of agencies selecting among regulatory alternatives — not a review of the full population of 28 economically significant regulations. The 236-proposed and 28-economically-significant counts both cover the same Jan. 2017–Mar. 2026 window, but the report does not explicitly confirm the 28 are a strict subset of the 236; see the computed note on the volume chart.
Sources(1) ▾
- U.S. Government Accountability Office, Tax Regulations: Opportunities Exist to Improve Economic Analysis and Taxpayer Engagement (GAO-26-108115) (2026-07-21) — the report's own summary: how many tax regulations Treasury/ proposed and finalized 2017-2026, how many crossed the $100 million 'economically significant' threshold, which -recommended economic-analysis practices Treasury and did and did not consistently follow, the public-comment volume and AI-generated-comment problem, the dial-in-only virtual hearing gap, and the six recommendations with Treasury's response gao.gov · original document
Comments
Always open. Logged-in readers can annotate paragraphs in place.
Between January 21, 2017, and March 31, 2026, Treasury and proposed 236 tax regulations and finalized 231⧉, including 28 designated "economically significant" — carrying $100 million or more in economic effects. A Government Accountability Office review⧉ of that record found the two agencies consistently did one thing right for those 28 high-stakes rules — analyzing alternative ways to design them — but inconsistently did the thing says matters most: producing specific cost, benefit, and revenue estimates to inform which alternative to pick. also found a rulemaking process straining under its own popularity, with some proposed regulations drawing tens of thousands of public comments, a rising share AI-generated, that has no policy to sort. issued six recommendations; Treasury agreed with two and rejected four.