A living dataset of published AI policies across higher education in 27 countries, classified by stance, maturity, and audience. We monitor what institutions put in writing, so leaders can see how the sector is actually responding. Updated Jun 1, 2026.
First-adoption year for 458 dated policies.
845 classified policies, by stance.
Share of each type with a verified AI policy.
Share of classified policies that explicitly require human oversight.
Number of institutions that name each tool.
All 6,602 monitored institutions, by policy maturity.
What the stances mean: Conditional — allowed with guardrails · Unclear — no clear position · Encouraged — actively promoted · Prohibitive — restricted or banned · Permissive — broadly allowed
29 active monitoring days, 16,183 coverage updates logged. Most recent first.
Method: institutions are drawn from public directories, AI policy pages discovered via search, verified by crawl, and archived in monthly snapshots. Snapshots are classified by an LLM on a locked rubric for stance, framing, and audience. Percentages use classified policies as the denominator unless noted.