We audited every institution in the 2025 Carnegie R1 classification — all 187 of America's highest-research-activity universities — for one thing: a published, university-wide policy on AI use, verifiable on the university's own official domain. The result is close to unanimity.
The census in one table
| Finding | Universities | Share |
|---|---|---|
| Publish an AI policy (YES) | 178 | 95.2% |
| Partial or unverifiable (UNCLEAR) | 8 | 4.3% |
| Confirmed no policy (NO) | 1 | 0.5% |
Source: US R1 AI Policy Audit 2026 (187 universities). Percentages rounded.
What “no” means at Wyoming
“No” here has a precise meaning, and it is worth stating exactly: three English-language searches found no adopted AI use policy on uwyo.edu. What exists instead is a 2023 working-group recommendations report containing draft syllabus language, library LibGuides on AI, and news about an AI commission and an enterprise platform. That is a university talking about AI governance. It is not a university that has adopted one.
This is not an accusation. Wyoming's working-group report shows the conversation is happening; the gap is adoption and publication. And the distinction matters for students: a draft syllabus paragraph and a library guide do not tell a student facing an academic-integrity hearing which rules applied to them.
The eight in-between
The eight UNCLEAR cases are more instructive than the single NO, because each one shows a different way to almost have a policy:
- Brown: a committee report, sample syllabus statements, and a policy explorer — but university-wide rules still being drafted.
- Clemson: substantive guidelines exist, but they surfaced only via search-engine crawl; direct page fetches were blocked, so the URL could not be confirmed.
- UT Health San Antonio: an official AI notice exists, but it is hosted on powerdms.com, not the university's own domain.
- UT Southwestern: a public page references “AI Responsible Usage” guidelines, but they sit behind VPN/intranet access.
- UMass-Boston: only a Writing Center-level policy, which defers to individual instructors.
- Miami: only program-level items — business school admissions, a public health handbook, law syllabi.
- UNH: IRB guidance for research plus a referenced system-wide standard, but no confirmable UNH policy page.
- Weill Cornell Medicine: only a library guide on a third-party host, plus event-specific abstract rules.
The pattern: every one of the eight has real AI activity. None of it adds up to an adopted, public, university-wide policy on its own domain. For a student, a researcher, or a journalist, that distinction is the whole ballgame — guidance you cannot open, or that lives on someone else's platform, is guidance you cannot rely on.
Why the holdout matters
Two years ago, the interesting question was whether universities would govern AI at all. At 95.2%, that question is answered. The live questions are different now: what do these 178 policies actually require — disclosure, detection, research rules? Who enforces them? And how much daylight is there between a one-paragraph statement and a real governance regime?
That is also the honest limit of this audit: it measures publication, not quality or enforcement. A published policy can be thin; an unpublished practice can be strict. Wyoming is the last holdout on publication. Whether it is also a holdout on practice is a question only its faculty and students can answer — and the fact that no public document lets outsiders check is, itself, the finding.
This census is the baseline for our Global AI Education Readiness Index: governance is one of its three dimensions, and a sector at 95.2% adoption is a sector ready to be ranked on what its policies contain, not just whether they exist.
— Study Intelligence, from the US R1 AI Policy Audit 2026. Data and methods open under CC-BY-4.0.
Study Intelligence is a research initiative by ACLAS College.