The biggest study of student AI use ever conducted just landed in Science, New York started vetting AI for over a million kids, and the research pile saying AI tutors beat the lecture hall keeps growing. Here are the six stories from the AI-in-education world that actually matter today — and what each one means for students, teachers and parents.
The biggest AI cheating study ever just dropped — and it's a warning
Berkeley News / Science · Published May 21, 2026
Researchers at UC Berkeley, Cornell and the University of Technology Sydney surveyed more than 95,000 undergraduates across 20 research universities — the largest study of generative AI use in higher education to date, now published in Science. Two-thirds of students said they use AI, and nearly 40% use it at least monthly.
The headline finding is a slippery slope: among daily AI users, 26% admitted using it to cheat, versus just 7% of monthly users. The more students lean on AI, the more likely the line between "help" and "cheating" quietly dissolves — partly because AI is now baked into search bars and writing tools, one click away from doing the whole job for you.
Bans won't save us. Smarter assessment will — and that's a conversation every school needs to start this term.
New York just put AI on probation for 1.1 million students
NYC Department of Education · June 2026
The New York City Department of Education released its preliminary AI guidance for the largest school system in the United States, covering 1.1 million students. The framework is strict: every AI tool must pass a vetting process before it can be deployed in a single classroom.
That vetting isn't a rubber stamp. Tools are evaluated for algorithmic bias, equity impact and developmental appropriateness — a direct attempt to stop districts from quietly rolling out chatbots that haven't been checked for who they might disadvantage. A fuller playbook is expected later in 2026.
Finally, a district treating AI tools like the high-stakes decisions they actually are.
OpenAI is going all-in on the classroom — and the numbers are big
OpenAI / The Edu Prompt · June 2026
OpenAI launched "The Edu Prompt," a newsletter aimed squarely at teachers and institutions, sharing product updates and practical classroom ideas a couple of times a month. The first issue spotlights a Duke University AI literacy assignment and the company's "Education for Countries" programme.
That programme is no longer a pilot: it has now reached over one million students in Jordan and more than 20,000 in Estonia. This is national-scale deployment of AI literacy training, run in partnership with governments rather than individual schools.
When the toolmaker also teaches the lesson, who's checking the homework? Worth watching closely.
ChatGPT for Excel and Google Sheets reaches schools worldwide
OpenAI · June 2026
ChatGPT for Excel and Google Sheets is now available globally for Enterprise, Edu and K-12 workspaces, with a free preview window for education customers. In practice that means teachers can pull AI directly into the spreadsheets they already use for grades, attendance and data.
It's a small-sounding update with a big classroom payoff: automated marksheets, instant data summaries and lesson-planning workflows that live where teachers already work, instead of in yet another tab.
This is the unglamorous automation that actually gives teachers their weekends back.
Universities are quietly giving up on AI bans
Inside Higher Ed · 2026
A new study finds faculty are moving away from outright bans on AI across nearly every discipline — the arts and humanities being the main holdouts. The hard, restrictive policies written in a panic after ChatGPT's 2022 launch are being replaced with something more nuanced.
Instead of banning, leading schools are adding manual review processes, student appeal systems, and bias audits of AI detectors. Some, like Vanderbilt, have paused AI-detection tools entirely over equity concerns — because the detectors flag innocent students too often.
Detectors were never going to win a cat-and-mouse game against AI humanizers. Good riddance.
The research keeps saying AI tutors beat the lecture hall
Nature, Scientific Reports · 2026
A randomised controlled trial published in Scientific Reports found that students using a research-based AI tutor at home learned significantly more than peers covering the same material in an in-class active-learning session — and they spent less time doing it.
It echoes a broader 2026 evidence base: a systematic review of 28 studies covering 4,597 K-12 students found AI tutoring systems produce statistically significant learning gains over conventional methods. Global spending on intelligent tutoring is now projected to hit $3.2 billion this year.
The catch nobody mentions: a tutor that never builds a relationship with you can only take learning so far.
🎬 Today's Top Pick for Video
The Biggest AI Cheating Study Ever Just Dropped
Story 01 is the one to film. It's specific, it's backed by a peer-reviewed Science paper, and it taps the exact anxiety every student, parent and teacher is feeling right now — with a counter-intuitive twist (bans don't work) that makes people click.
"95,000 students. 20 universities. The biggest study of AI in education ever done just confirmed every teacher's worst fear — and it's not the students you think who are most at risk."
📚 Revise smarter with RevisionLab
AI-powered study tools, personalised revision plans and exam-technique drills built for GCSE and A-Level students who want results without the burnout.
Explore RevisionLab →Prefer something to hold? Browse top-rated GCSE & A-Level revision guides on Amazon.
As an Amazon Associate I earn from qualifying purchases.
Get this digest in your inbox 📬
Five minutes a day on how AI is really changing education — no hype, no overwhelm.
Join the newsletter →