Four things to know
before your first meeting
The table of contents for a busy week — read this, then decide whether to keep going.
Admissions AI policies found at most institutions queried. A watchdog's public-records investigation of 24 public universities found AI moving into admissions offices with almost no written rules — while the same campuses demand AI disclosure from students.
OpenAI joins the giveaway. Four free months of ChatGPT Plus for verified U.S. college students — every major lab now subsidizes student premium AI this semester.
Detection's second generation. Nature reports the new detectors are far more accurate than the tools faculty learned to distrust — and the University of Chicago is using one on coursework.
Teacher prep gets a framework. AACTE published the first national framework for teaching AI inside university teacher-preparation programs — moving the question from district procurement into colleges of education.
The accountability gap
moved upstairs
Four stories with details, sources, and what each one means for your campus.
Universities are policing student AI use while leaving their own admissions offices ungoverned
Student Defense's SHAPE AI Initiative filed public records requests with 24 public colleges and universities in spring 2026 asking for policies governing AI use in undergraduate admissions. The report, released August 30, found most institutions have none — even as admissions offices adopt AI for reading applications the report calls "high-stakes, individualized, and legally sensitive."
Chief Counsel Dan Zibel: "too many colleges and universities have yet to establish the policies and trainings needed to ensure AI is used responsibly in admissions offices." The report recommends human accountability for decisions, disclosure to applicants, a bar on undisclosed staff AI use, and applicant-data protection. The gap mirrors student-facing integrity policy in reverse: institutions demand disclosure from students that they do not practice institutionally.
Student Defense, Aug 30 →If your institution requires students to disclose AI use, expect the same question to be turned back on admissions, advising, and grading — and it will come from journalists or lawyers before it comes from faculty senate.
Update on the back-to-school giveaways (Aug 22 issue): OpenAI adds four free months of ChatGPT Plus
After Google's free year for students, OpenAI opened four free monthly billing periods of Plus — an $80 value — for verified U.S. college students, claimable through October 31 via SheerID. Marc Watkins (University of Mississippi) notes students now hold simultaneous free premium trials from OpenAI, xAI, and Google, calls the equity effects "a staggering blow," and warns that assessments vetted against free-tier ChatGPT no longer describe what students actually have. Assume premium AI in your classroom until at least December.
OpenAI, Aug 25–26 → Watkins, Aug 30 →
Nature: AI detectors quietly got much better — and Chicago is using one on coursework
A Nature news feature reports the new detector generation — Pangram most prominently — correctly identifies human writing as human "almost all the time" in independent analyses, a reversal of the false-positive problem. The University of Chicago has started using Pangram to vet coursework; alphaXiv shows verdicts on arXiv preprints. Worth noting: this cuts against MIT's advice (Aug 26 issue) not to rely on detectors, and peer-reviewed accuracy data across diverse student writing does not yet exist.
Nature feature → Nature letter →
Teacher-education programs get their first national AI framework
AACTE released a framework for how university teacher-preparation programs should teach AI — technical skills, pedagogical knowledge, and ethical judgment — developed by a task force of members and outside experts. Until now most teacher AI training came from districts and vendors, often platform-specific; co-chair Peña Bedesem (Kent State) and Anne Tapp Jaksa (Michigan) argue AI now touches the fundamentals candidates learn: planning, assessment, differentiation, ethics. If your institution prepares teachers, this is the document your accreditation and curriculum committees will be handed this year.
Education Week, Aug 27 →What the studies say,
what colleagues tried
Two findings and two classroom moves worth borrowing.
From the Research
"It's here and we can't fight it" — engineering faculty are improvising without guidance
A Frontiers in Education study finds adoption sharply uneven: some faculty redesign assignments around AI while same-department colleagues prohibit it — and both report the same root condition: policy, training, and pedagogy haven't kept pace. The takeaway is an argument for department-level norms: the inconsistency students complain about lives between adjacent classrooms.
Frontiers, Aug 28 →What universities are for when AI makes knowledge cheap
A conceptual review proposes the "Human-Agency Infrastructure Framework": when explanations become abundant, the scarce goods shift to judgment, trust, credible evidence, belonging, and ethical agency. Useful curriculum-committee frame: ask which scarce goods each course actually builds, not just what students may use AI for.
MDPI, Sep 1 →How Others Are Doing It
He swapped his exam for an essay on why students cheat
Tipped off that students planned to use AI on a take-home exam, Grant replaced it with one essay: use the organizational psychology you've studied to design a way to motivate students not to cheat. His survey of 95 students first: hardly any saw cheating as an integrity violation. His conclusion — obstacles don't work; change motivation, not surveillance. Borrow the move: make the integrity problem itself the assessed content.
NYT, Aug 26 →A hidden prompt measured unedited AI use: 32 of 35 students
Gibson embedded hidden instructions in a writing prompt — triggered only if students pasted it into a chatbot and submitted the output unread. 91% turned in gibberish about Madagascar. The diagnostic matters more than the trap: the problem is not AI use but zero editorial oversight — a teachable distinction. (Primary post not independently located; details as reported.)
Write-up, Aug 28 →Seven fast items,
two dates, three actions
Scannable in under a minute — plus what to raise at your next department meeting.
- 1Google and Khan Academy ship Gemini-built Khanmigo upgradesInteractive diagrams + teacher-controlled practice builder · Aug 27
- 2Anthropic makes Claude for Teachers a free enterprise offering for schools and districtsSSO, role controls, FERPA-aligned K-12 terms · Week of Aug 28
- 3GPTZero: only 20% of educators report a consistently applied school-wide AI policy32% say the hard part is defining acceptable support vs. cheating · Aug 28
- 4Every Learner Everywhere: "Where AI Meets Accessibility"Reframing AI in higher ed beyond integrity debates · Aug 26
- 5The Chronicle profiles Ethan and Lilach Mollick as AI "evangelists"Institutions should lead, not wait for vendors · Aug 28
- 6The Atlantic: "A Society of Cheats"Crimson survey: a third of Harvard grads used AI against permission · Aug 26
- 7The Scholarly Kitchen: readers now accuse posts of being AI-written dailyEscalating complaints + AI-labeling experiments · Aug 27