Every published Top AI Stories item tagged with Education, newest first.
The OECD published its Programme for International Student Assessment, known as PISA, on September 8, covering more than 760,000 fifteen-year-olds who represent about 33 million students across 91 countries. It is the first round run since AI use became ordinary in schools. After adjusting for socioeconomic status, students who never use AI generally outperformed those who do in science. The detail matters more than the headline. Using AI to summarize a text you were assigned to read, or to draft a writing assignment, tracked worst; using it for preliminary research or general help learning tracked far better. Frequency was a curve rather than a line, with yearly and daily users faring worst and monthly or weekly users best. And students who had been taught to judge the quality of what an AI tells them outperformed non-users outright. Reading and mathematics scores have fallen sharply across the OECD over the past decade, a longer trend than AI, but the report lands on classroom policy at a moment when districts are already legislating.
Mayor Zohran Mamdani announced a one-year moratorium barring classroom AI use through eighth grade for the 2026 to 2027 school year, covering roughly 600,000 students. Chatbots built for companionship are banned in every grade, and teachers may not use AI to grade work. High schoolers get limited access plus AI literacy classes twice a year, and a pilot in up to five classrooms will test vetted tools including Quill, Edia and Brisk Teaching. District leaders call it among the most restrictive policies in the country.
HBS Foundry, an eight-week program for entrepreneurs, pairs weekly live instructor sessions with AI avatars of those same instructors, built by HeyGen, that give feedback during practice pitches and mock board meetings. Project director Katharina Rings said the avatars replaced her original chatbot idea after students asked for something more guided. Instructor Jeff Bussgang called his digital copy "creepy" and added, "My students love it." A New York Times reporter who tried it noted the avatar's smile stayed noticeably frozen through her pitch.
Google added a batch of study features across Search and Gemini: AI-generated interactive visuals and 3D simulations, custom practice quizzes, full-length SAT practice tests, a Lens mode that explains a photographed worksheet, and a student hub holding notebooks, flashcards and quizzes. The sharper move is commercial. Eligible US college students get **a year of Google AI Pro at no cost**, a plan that normally runs $200 a year, and it auto-renews at full price when the year ends. Students outside the US get the cheaper AI Plus tier instead.
ChatGPT for Teens began rolling out globally, and teenagers do not opt in — anyone who states an age of 13 to 17, or whom OpenAI's age-prediction system estimates is under 18, is moved over automatically. The model will not use romantic language, imply feelings, call itself a friend, or encourage emotional dependence, and a Study Hours setting can force Study Mode on a schedule. OpenAI has not published how accurate its age prediction is, and Common Sense Media maintains that social AI companions are not safe for children at all.
India's address from the Red Fort was carried live in 22 Indian languages through Bhashini, the government's open-source translation stack — the first time the annual speech has run through an AI pipeline at national scale. In the same address, Prime Minister Narendra Modi pledged AI skills training for 10 million young Indians within a year. That target deserves scrutiny rather than applause: the government has published no implementation detail, and the existing national skilling scheme has certified only tens of thousands.
Researchers built a corpus of 88 billion tokens filtered to the US elementary-school curriculum, excluding anything taught above fifth grade, then trained a model with 5 billion parameters on it. Scaling, reinforcement-based post-training and in-context learning all made the model better at using what the curriculum had taught it — and none meaningfully lifted performance outside that boundary. It is a clean demonstration that pretraining data sets a ceiling later techniques mostly cannot raise.
The National Autonomous University of Mexico, known as UNAM, ran its entrance exam fully online for the first time, using a locked-down browser and behavior-monitoring software from Territorium Life that watched webcam feeds for wandering eyes, extra people in the room and background conversation. The scores came back nothing like history — about 16 percent of test takers cleared 100 correct answers out of 120, against roughly 3 percent from 2021 through 2025. A technical commission called for an in-person control exam, and roughly 58,000 applicants, about 36 percent of the field, must sit it again under human supervision.
Google DeepMind and India's Atal Innovation Mission launched ATL Saathi, a Gemini-powered web app that helps teachers in the country's network of school "tinkering labs" plan projects, deliver micro-lessons, and mentor students across eight languages. The tool is starting as a live pilot in an initial cohort of 100 schools, aimed at cutting educators' administrative load. It is early — no outcome data yet — but it is a concrete deployment built to stretch scarce hands-on STEM-teaching capacity.
Norway will bar generative AI for students in grades one through seven — ages six to 13 — and restrict it for older pupils, with teens 14 to 16 allowed to use it only under teacher supervision. Prime Minister Jonas Gahr Store said the rules, which take effect when the school year begins in late August, aim to stop young children from skipping essential steps in learning. The move follows a 2024 Norwegian smartphone ban and a broad decline in national test scores.
Failing grades jumped sharply in several UC Berkeley computer-science courses this spring — 35 percent of students failed the introductory CS 10 class, up from under 10 percent in prior years — and teaching professors point to over-reliance on chatbots like ChatGPT, Claude, and Gemini as the primary driver. Faculty say students lean on the tools for take-home work, then arrive at in-person exams unprepared, and that incoming students increasingly lack linear-algebra and proof skills. More than 1,300 UC faculty are now petitioning to reinstate standardized-test requirements for STEM admissions.
Malta and OpenAI announced a partnership making the island nation the first country to offer free ChatGPT Plus to every citizen and registered resident. Eligibility requires completing AI for All, an online AI-literacy course developed by the University of Malta, with the Malta Digital Innovation Authority managing distribution. Subscriptions run for one year starting in May 2026; the deal is the first under OpenAI's OpenAI for Countries initiative, which aims to anchor national AI adoption around ChatGPT.
Anthropic announced a four-year, $200 million commitment to the Gates Foundation, structured as a mix of cash grants, Claude usage credits, and technical support from its Beneficial Deployments team. Focus areas span global health (polio, HPV, and preeclampsia and eclampsia among named disease targets), life sciences, education, and economic mobility, with regional emphasis on sub-Saharan Africa, India, and other low- and middle-income countries. Launch partners include the Institute for Disease Modeling and the Global AI for Learning Alliance. It is the largest single philanthropic commitment from a frontier lab to date and a meaningful structural signal that Anthropic intends to be measured by social-impact deployments alongside its commercial book.
Princeton faculty voted — with a single dissenting vote — to mandate proctoring for all in-person exams starting July 1, 2026, ending a 133-year tradition of unproctored testing under the student honor code. The proposal cites "AI and personal electronic devices as major catalysts," noting that AI tools on small devices make cheating harder for peers to observe, which has hollowed out the peer-reporting model the honor system relied on. The Daily Princetonian's 2025 Senior Survey of over 500 graduating seniors found 29.9% admitted cheating on an assignment or exam, 44.6% said they knew of violations they did not report, and only 0.4% said they had reported a peer. Instructors will now serve as exam witnesses but will not actively monitor.
The Walton Family Foundation's sixth Gen Z AI poll, fielded with Gallup across 1,572 respondents aged 14 to 29, finds weekly AI use stuck at 51% — growing only four points year-over-year — while anger toward AI jumped to 31% from 22%. Eighty percent now say relying on AI to finish tasks faster will make learning harder, and fewer than 20% would choose AI for tutoring, financial advice, or customer service. Schools tightened in parallel: 74% of Gen Z students report a school AI policy, up 23 points from the prior year.
In a new paper on arXiv, researcher Valerio Capraro coins *LLMorphism* — the biased belief that human cognition operates the way a large language model does. He argues the framing flows from two channels: analogical transfer (projecting LLM features onto humans) and metaphorical availability (LLM vocabulary becoming the dominant way to talk about thought). Capraro: "the issue is not only whether we are attributing too much mind to machines, but also whether we are beginning to attribute too little mind to humans." The paper traces implications across work, education, healthcare, and human dignity.