For University Faculty & Staff · ~10 min read · July 29 – August 4, 2026 Newsletter Home
Intelligence Briefing · No. 13

AI in
Education

The sector retreats from unsupervised assessment, Mexico's flagship university reels from an AI-proctored admissions meltdown, and non-majors flood AI courses as computer science enrollment slides.

3,000
UNAM exams invalidated
133
Years of unsupervised Princeton exams, ended
77,000
Students in largest AI-tutor usage study
11%
Leaders who say AI delivered measurable value
Issue briefing video

Watch the Issue 13 briefing.

A short video companion to this AI in Education issue, placed here before the section navigation so readers can watch before diving into the full briefing.

This week
in 30 seconds

Five things worth knowing before you walk into your department. Items fully covered in the August 1 issue are omitted.

🎓 Higher Ed
133

The retreat from unsupervised assessment accelerates. University of Chicago Law bans laptops and phones for incoming students, and Princeton votes to end its 133-year tradition of unsupervised exams — the clearest signal yet of where assessment is heading.

🌍 Global
3,000

Mexico's UNAM invalidated roughly 3,000 admissions exams in an AI cheating-and-proctoring scandal that reached President Sheinbaum.

🎓 Higher Ed

Computer science enrollment is falling, but Northwestern doubled its CS teaching staff anyway — to serve non-majors flooding into AI courses, per AP.

📊 Research
77K

The largest study of real AI-tutor usage — 77,000 students in Germany — finds AI-supported learning woven into everyday study schedules, not replacing them.

🏛️ Policy
11%

Only 11% of higher-ed leaders strongly agree AI has delivered measurable value — yet 71% still name scaling AI their top priority.

Main stories

The developments most likely to change how your campus teaches, tests, and plans.

University students taking an invigilated, handwritten exam in a lecture hall Image: Nano Banana 2

The retreat from unsupervised assessment accelerates

The University of Chicago Law School has banned laptops and smartphones for incoming first-year students — all exams move in-class, without internet, files, or apps — while pairing the ban with a plan to teach effective AI use later in the program. Princeton, meanwhile, voted to end its 133-year honor-code tradition of unsupervised exams. Three-quarters of respondents to an AAC&U survey say they are concerned about generative AI cheating.

A July 20 HEPI report frames the alternative: "the pursuit of [integrity] through approaches that work," urging programs to move away from essay-based submissions rather than lean on detection. Backdrop, covered in the August 1 issue: the Yale GPTZero federal lawsuit, Brown's take-home midterm investigation, and Alcorn State's hidden-prompt trap, still topping TikTok feeds into August.

The Independent, July 28 →
What this means for your campus

Decide before syllabi lock which of your assessments move in person. The sector's direction of travel is now unambiguous — and detector-based enforcement is the legally weaker alternative. Chicago's ban-then-teach sequencing is a model worth copying: protect foundational skills first, teach the tool second.

Administrator reviewing a tall stack of printed exams Image: Nano Banana 2

UNAM's AI cheating scandal leaves thousands of applicants in limbo

Mexico's most prestigious university took its admissions exam online for the first time, guarded by AI surveillance — then invalidated ≈3,000 tests (2%) after a wave of suspiciously perfect scores. Investigators found cheating via phones, stand-ins, and identity theft; applicants who feel wrongly flagged now blame the AI proctor itself, and President Sheinbaum suggested involving the attorney general. High-stakes online testing concentrates both cheating risk and wrongful-flag risk.

CNN, August 2 →
Researcher working at a laptop in a university library Image: Nano Banana 2

CS enrollment falls — and AI courses overflow with everyone else

AP finds entry-level software hiring cooling and computer science enrollment declining, while demand for AI instruction from non-tech majors surges. Northwestern doubled its CS teaching staff, is adding an AI major, and is streamlining prerequisites for non-majors. Harvard's Paul LeBlanc calls the pace "remarkable for a sector infamous for moving slowly." If your department doesn't offer an accessible AI route, students will find one elsewhere.

AP, August 3 →

Research
& practice

What the data says, and what colleagues are actually doing with it.

From the Research

Large-scale usage data · Germany

77,000 students, zero surveys: the largest look at real AI-tutor use

What was studied: IU International University of Applied Sciences analyzed anonymized behavioral data — not self-reports — from ~77,000 online students using its AI learning companion Syntea, the largest empirical analysis of a dedicated AI learning assistant to date.

What they found: AI-supported learning is now part of everyday student life, and usage mirrors students' own schedules rather than replacing study time — a behavioral baseline self-report surveys have missed.

What you could do with this: If your institution runs an AI-tutor pilot, ask for actual usage logs rather than satisfaction surveys. The behavioral data is where adoption claims get tested.

IU News, July 31 →
Sector survey · United States

The AI value gap: big bets, thin returns

Across 75 higher-ed institutions surveyed, 71% name scaling AI their top priority for the year — but only 11% strongly agree AI has delivered measurable value so far. The gap between investment and demonstrated return is the number to bring to your next budget conversation.

Coastal Cloud, July 28 →

How Others Are Doing It

Cross-disciplinary teaching · Virginia

A music professor co-teaches AI with a computer scientist

Who: Benjamin Guerrero, assistant professor of music at Eastern Mennonite University.

What he's doing: Co-teaching a course next year with a computer scientist so music students can use AI tools for tasks like editing and generating drum tracks. Guerrero, who has advised the National Association for Music Education on AI, frames it as "no more disruptive than the record player, the radio, the metronome, the synthesizer or the computer."

What you could borrow: Cross-disciplinary co-teaching is a low-cost way to bring AI into a non-technical major without building a curriculum from scratch.

AP, August 3 →
Assessment design · Illinois

Chicago Law's ban-then-teach sequencing

The law school's device ban for first-years is paired with a deliberate plan to teach effective AI use later in the program — a sequencing model that protects foundational reasoning first and introduces the tool once judgment exists to evaluate its output.

The Independent, July 28 →

Quick hits
& radar

Worth knowing, no commentary needed — plus dates to watch.

On My Radar — Dates
August 10, 2026
OpenAI Student Collective applications close — flag to undergraduates who lead AI clubs on your campus.
August 14, 2026
Handshake × Google student contest closes — eight winners get $500 and a career conversation at Google, alongside a free year of Google AI Plus.
Try This Week

Convert one take-home assessment component to in-class, oral, or process-based before fall syllabi lock. Then draft one AI-permitted assignment with required disclosure, modeled on Chicago Law's ban-then-teach sequencing.