iTeachAI NEWS
Ten things in AI and education, every week
Edition 45 · Saturday, November 14, 2026
This week

Who Else Touches Your Data?

I teach gifted students, and the fastest way to lose them is to accept an answer they cannot explain. This week I found myself asking the same of the platforms my classroom runs on. One upgrade problem at Microsoft took Teams, Outlook and Copilot down around the world for more than a day. A breach at a company most Notion users had never heard of exposed their AI chats. Last year the Pentagon ended a Microsoft program after learning that the people supervising work on its cloud systems could not evaluate what they were supervising. The question for every school is the same one. Who is behind the tools we use, and who is checking their work?

1

Platform | Microsoft introduces Microsoft Classroom, a standalone home for teaching and learning

Microsoft Classroom pulls Assignments, Classwork, Grades, OneNote Class Notebook, Learning Accelerators, Insights and Reflect into one place, with the AI planning workspace Teach built in. It does not replace Teams for Education. Microsoft says it will be included at no additional cost in current A1, A3, A5 and A7 licenses starting next year, with a preview planned for early 2027.

Why it matters: If your district runs on Microsoft, the question for next fall is no longer which add-on to buy. It is whether the gradebook, the assignments and the AI planner all live in the one tool your teachers already have. Sign up for the preview before the decision gets made for you.

2

Higher Ed | England's university regulator says AI has made coursework close to impossible to assess

The Office for Students and Advance HE spent four months talking with about 300 students and staff at more than 120 institutions. Their report, Making AI Work for Students, says unsupervised essays and open-book tasks can be finished with AI in minutes without detection. The regulator wants far more AI-free assessment, including in-person exams, presentations and oral exams, and says institutions that cannot assess reliably could face sanctions.

Why it matters: This is the first regulator to say out loud what teachers have known for two years. Detection tools are not the answer. The assignment is. If a task can be done in four minutes by a chatbot, it was measuring the wrong thing.

3

Research | Harvard builds an age-by-age guide to AI from more than 5,000 pieces of research

The Harvard Graduate School of Education launched the Compass Collaborative on November 10 and released its first tool, the AI Guidance by Age Matrix. You choose your role, a developmental area and an age range, and it summarizes what the research says AI may help or hinder. For 9 to 11 year olds, for example, it says adaptive practice can help while AI homework and answer tools may undermine the retrieval that builds memory.

Why it matters: Most AI guidance treats a third grader and a high school junior as the same user. This one starts with the child's age and works outward, and it is free. Bring it to your next grade-level meeting.

Source: Education Week →
4

Research | Stanford review of 52 studies: AI is helping teachers more than it is helping students

The AI Hub for Education at Stanford's SCALE Initiative reviewed 52 rigorous studies, 32 of them new. Students who used AI to write essays or solve problems learned little and did worse later without it. Teachers fared better: those who used AI for a first pass at grading gave more detailed feedback, and lesson planning took less time without losing quality. One caution: students who believed a lesson was AI-generated reported lower motivation.

Why it matters: The evidence is pointing in one direction. AI in the teacher's hands, yes. AI doing the student's thinking, no. Build your classroom rules around that line and you will be ahead of most policy documents.

Source: Chalkbeat →
5

Infrastructure | A global Microsoft outage takes down Teams, Outlook and Copilot for more than a day

The disruption began on November 12 and ran into November 13. Microsoft first blamed elevated load on its infrastructure in multiple regions, then traced it to a problem during a network capacity upgrade. Executive Charles Lamanna said there was no indication of malicious or cyber activity. OpenAI and Anthropic both reported problems with their Microsoft 365 connections.

Why it matters: When the platform goes down, the lesson plan, the gradebook and the parent email go down with it. Microsoft says this was a technical problem during a network upgrade, not an attack. That is the point. One company's mistake reached classrooms on every continent. Every teacher needs one day of instruction that works with the Wi-Fi off. If you do not have it printed, print it this week.

Source: Malay Mail →
6

Models | Google ships two Gemini 4 models in one week, Carbon and Hydrogen

Gemini 4 Carbon arrived on November 10 as Google's most capable model, aimed first at Ultra subscribers and developers. Two days later came Gemini 4 Hydrogen, a faster and cheaper model for everyday tasks that Google is putting into the Gemini app and into Google Search for all users globally.

Why it matters: Hydrogen is the one your students will meet, because it sits inside Search. If your district is still deciding whether to turn Gemini on, know that the model behind the search box just changed again.

Source: Google →
7

Models | Anthropic releases Claude Fable 5.2, faster and cheaper than the version before it

Claude Fable 5.2 launched November 9 with gains in coding, reasoning and vision. Anthropic says output is up to 30 percent faster than Fable 5.1, at a lower price for developers. The company still recommends Claude Opus 5.5 as the everyday model for most coding and knowledge work.

Why it matters: The top models are now getting cheaper and faster every two months. That is good news for the education tools built on them, and a reminder that any AI policy naming a specific model is out of date before it is approved.

Source: 9to5Mac →
8

Agents | United Airlines bans AI agents from booking flights

United said on November 11 that it is blocking unauthorized third-party AI assistants from making reservations, on its own site and through travel agencies and aggregators. It cited exposed payment details and login credentials, unintended purchases, and federal rules that require a traveler to confirm certain things in person. Southwest has gone the other way and opened its flights to ChatGPT users.

Why it matters: This is the clearest classroom example yet of the difference between AI that advises and AI that acts. Ask your students: who is responsible when the assistant buys the wrong ticket? That is a whole lesson on agency and consent.

Source: CBS News →
9

Privacy | A breach at an AI vendor exposes Notion AI chat transcripts

Fireworks AI, a company that runs AI models for other products, disclosed a second security incident. Notion confirmed that chat transcripts from its Insights feature were in the exposed data, affecting fewer than 1 percent of its workspaces. Notion has cut ties with the vendor and turned the feature off.

Why it matters: Your students' data does not only live with the tool on the screen. It also lives with the companies behind that tool, and with the companies behind those. Before you approve any AI product, ask the vendor two questions. Who else touches what my students type? And who checks their work?

Source: eSecurity Planet →
10

Science | Google DeepMind releases AlphaGenome 2 to predict how DNA changes affect specific cells

Announced November 11, AlphaGenome 2 was trained on more than 100 million individual cell profiles across 68 tissues. Instead of averaging across a whole tissue, it predicts the effect of a DNA change in one cell type at a time. It is available now for non-commercial research.

Why it matters: Biology teachers, this is your hook for the next genetics unit. The same kind of AI your students use to write is being used to read the genome, one cell type at a time.

Source: Google DeepMind →
Four More Worth Your Time
11

K-12 | California teachers ask for a three-year pause on AI in K-8 classrooms

The California Federation of Teachers, which represents 120,000 educators and faculty, sent a letter asking the state to pause AI in K-8 public school classrooms for three years. The union says educators are not prepared to supervise AI use and that students struggle to tell which tools are safe. It cited research including a study in which students who used GPT-4 scored 17 percent lower on a later unassisted math test.

Why it matters: You do not have to agree with a pause to hear what is underneath it. Teachers are saying they were handed tools without training. The fix for that is professional learning, and it is cheaper than a three-year delay.

Source: GovTech →
12

Industry | Anthropic sets terms for one of the largest stock market debuts ever

Anthropic, the company behind Claude, set a price range of $70 to $78 a share for 825 million shares, to trade on the Nasdaq under the ticker ANT. Pricing is expected November 18 with trading the next day.

Why it matters: The companies building classroom AI are about to answer to public shareholders. Watch what happens to free educator plans and student pricing over the next year.

13

Media | Four AI companies back a simple way for websites to opt out of AI training

Anthropic, Google, Microsoft and OpenAI have agreed to support a proposed "policy token," a line of text a publisher adds to its robots.txt file to say its content may not be used to train AI. Critics note that honoring it is voluntary.

Why it matters: This is a ready-made media literacy lesson. Who owns what is on the internet, and who gets to say no? Have students look up the robots.txt file of their favorite site.

Source: Digiday →
14

Research | Stanford sets out to build an AI model of the human brain

The Wu Tsai Neurosciences Institute launched the Stanford Digital Brain Project, a five-year effort across Stanford's schools to collect detailed brain imaging data and use it to train models that predict human brain activity. The hope is to change how brain disorders are understood and treated.

Why it matters: We spent this whole edition asking what AI does to learning. Here is the reverse: AI being used to understand the organ that does the learning.

Source: The Stanford Daily →
Dr. Janette Camacho, Ed.D. Founder and CEO, iTeachAI Academy