The week in
five headlines
If you read nothing else this week, read these.
The largest study ever of undergraduate AI use — 95,513 students — was published in Science, finding that cheating climbs steadily the more often students use AI, and that bans are the wrong fix.
AI became commencement season’s most-booed topic, with Class of 2026 graduates jeering speakers at Arizona, UCF, and Middle Tennessee State who told them to embrace it.
EDUCAUSE’s 2026 Horizon Report names four new “signals of change” for classroom AI — AI learning assistants, AI-powered textbooks, 3D model generation, and apprenticeship-for-credit models.
A study of 96 UK universities found two in five have no AI policy a student can easily find — and most of the rest read as detection-and-discipline.
A surge of legislation across more than two dozen states, plus a summer teacher-training push, is driving AI literacy into K-12 — reshaping the students headed to your campus.
Five stories that
moved the needle
The developments most likely to reach your department this week.
The largest study yet of student AI use lands in Science
A survey of 95,513 undergraduates at 20 public research universities — the largest of its kind — was published May 21 in Science by researchers at UC Berkeley, Cornell, and the University of Technology Sydney. About a third of students regularly use generative AI for coursework, and 9% have used it to cheat.
Misuse rises with frequency: 26% of daily AI users reported cheating, versus 7% of monthly users. Use was higher in data-intensive disciplines and lower among female students (33% vs. 45% of men) and underrepresented-minority students (29% vs. 39%) — a measurable access gap. The data was collected during the 2023–24 academic year; co-author Igor Chirikov calls it conservative and says it already feels “from a past life,” so treat 9% as a floor, not a ceiling.
The authors’ prescription is pointed: discipline-specific assessment reform — not blanket bans, and not universal AI-detection regimes.
Read the full report →You now have peer-reviewed evidence to move assessment-redesign conversations to the department level — and to push back on both the blanket ban and the detection arms race. The study’s own framing makes the stakes plain: when AI inflates grades, it is the credibility of your credentials that erodes.
AI becomes commencement season’s most-booed topic
Graduates booed former Google CEO Eric Schmidt at the University of Arizona, executive Gloria Caulfield at UCF (for calling AI “the next industrial revolution”), and music executive Scott Borchetta at Middle Tennessee State, who replied, “Deal with it.” It is not a rejection of the tools — 57% of students use AI weekly — but anxiety: two-thirds are pessimistic about the job market and roughly 4 in 10 have considered changing their major because of AI.
What it means for your campusStudents are anxious, not just curious — frame AI conversations around their job prospects and agency, not only productivity.
GovTech →2026 EDUCAUSE Horizon Report maps the next wave of classroom AI
EDUCAUSE released its 2026 Teaching and Learning Horizon Report, adding a new “signals of change” section — AI learning assistants, AI-powered textbooks, 3D model generation, and apprenticeship-for-credit models. Built on the STEEP framework, it frames a sector-wide shift “toward transparency and invitation over surveillance and detection.”
What it means for your campusA free, citable planning document to put in front of a curriculum committee or instructional-design team before fall.
EDUCAUSE →A study of 96 UK universities finds AI policy is mostly policing
HEPI’s Policy Note 71, by Edinburgh Napier University professor Sam Illingworth, found two in five UK universities have no AI policy a student can easily find online — and most of the rest “use the language of learning but actually operate as detection-and-discipline frameworks.” It names four exemplars — Stirling, Canterbury Christ Church, Arts University Plymouth, and Durham — that focus on assessment design over detection.
What it means for your campusA practical prompt to pull up your own department’s AI policy and ask which model it follows — and whether a student could even find it.
HEPI →State legislatures move fast on classroom AI
FutureEd’s legislative tracker shows dozens of bills across more than two dozen states addressing AI in instruction this session — from mandated state guidance to local “AI coordinators.” In parallel, the Computer Science Teachers Association is running an $11M multi-state summer training push, and a White House Task Force on AI Education has made K-12 AI a federal grant priority.
What it means for your campusStudents arriving over the next few years will increasingly have had formal AI instruction — and your education school’s teacher-prep curriculum is now a competitive differentiator.
FutureEd →Evidence, and
what to do with it
Two studies worth knowing, and two campuses worth copying.
From the Research
Universities need a change model built for uncertainty
University of Colorado Boulder physicists David Perl-Nussbaum and Noah Finkelstein argue the usual model of educational change — test, build evidence, then roll out — no longer fits AI, because students adopt tools before any evidence base exists. They reject both blanket bans and uncritical adoption, and call for treating AI changes as small, reversible experiments with evaluation built in.
EdTech Innovation Hub →Students use AI more — and trust it less
RAND surveyed 1,214 people aged 12–29 on its American Youth Panel. AI homework use rose from 48% to 62% in seven months; over the same window, the share saying AI harms critical thinking climbed from 54% to 67% — 75% among female students versus 59% of males. Their ambivalence is an opening: assign a short reflection asking where AI helped them learn versus only finish.
RAND →How Others Are Doing It
Utah hands AI literacy to students as co-designers
The University of Utah built a Student-Led AI Symposium and a recurring weekly “AI Tinker Lab” — collaborative spaces for building AI literacy rather than policing it. The 2026 EDUCAUSE Horizon Report cited the symposium as a national example of shifting “toward transparency and invitation over surveillance and detection.” Borrow the idea: a low-cost tinker lab or symposium signals invitation, not surveillance.
University of Utah →Georgia State runs a 21-faculty generative-AI pilot
Across spring 2026, 21 faculty across eight colleges at Georgia State tested 18 approaches to integrating generative AI in courses reaching about 1,100 students, all on one FERPA-compliant, research-safe platform — deliberately low-stakes, with shared instructional-design support. Borrow the idea: a small, opt-in faculty cohort on a common platform moves people further than a top-down mandate.
Georgia State CETLOE →Everything else
worth a glance
Six fast headlines, the dates ahead, and one thing to try.
-
1OpenAI adds Singapore to its “Education for Countries” program At the Education World Forum, OpenAI welcomed Singapore and launched “OpenAI Luminaries,” an educator co-design track.
-
2AI educator Andrej Karpathy joins Anthropic The OpenAI co-founder, who built widely used free AI courses, joined Anthropic’s pre-training team and paused his AI-native school.
-
3A prize-winning short story is flagged as AI-written Wharton’s Ethan Mollick flagged a Commonwealth Short Story Prize winner published in Granta as likely AI-generated; one detector scored it 100%.
-
4An AI name-reader malfunctions at graduation Glendale Community College’s AI name-reading system skipped dozens of graduates’ names; Columbia faculty separately protested AI-voiced name-reading.
-
5Google publishes its own AI learning-impact studies Google reported two company-run studies (Sierra Leone, Italy) claiming Gemini-linked learning gains — notable, but vendor-run.
-
6AI skills are becoming a hiring bar — and some colleges lag As employers demand AI skills, schools including Ohio State and Purdue now require a basic level of AI competency to graduate.