Microsoft’s AI privacy standard for schools: what it covers

October 2, 2026

In September, Microsoft signed a set of AI privacy terms for American schools with the American Federation of Teachers and its New York City local, the UFT. From November 1, every district can write those terms into its Microsoft contract. Before anyone treats this as a clean bill of health for the AI in your room, read what was actually signed. Then notice how much it leaves out.

What the standard says, in plain terms

The National AI Safety & Privacy Standard was announced on September 9, 2026, by the AFT, the UFT and Microsoft. The privacy promise fits in one sentence of Microsoft’s own announcement: student and educator data “will not be used to train AI models, sold or repurposed,” and schools keep control over how that data is used, retained and deleted.

Training a model means feeding it text so the next version learns from it. In plain terms, what a student types into a covered Microsoft tool is not supposed to end up teaching the product. Education Week read the agreement and reports one narrow exception, for specific safety-related data.

The other two parts are safety and transparency. The AFT release says AI “cannot make decisions without human oversight,” and that systems must be designed to avoid harmful or manipulative experiences. Education Week reports that AI “companions,” chatbots built to simulate a relationship, are explicitly banned. Companies also have to tell schools clearly about product changes, risks and controls.

Here is what makes it more than a pledge: districts can write these protections into their Microsoft customer agreements, which makes them contractually enforceable. According to the AFT, a district that finds a breach can end the agreement and seek damages.

What changes on November 1, and what does not

According to Microsoft’s fact sheet and K-12 Dive’s report, the protections become available to every school district in the country on November 1, 2026. Districts can add them to new or existing Microsoft agreements without renegotiating or renewing.

Notice the wording: available, and can add. The official texts describe something a district does, not something that happens to it automatically. The press has not agreed on this point, and some reports describe the terms as applying to every existing contract. If you want to know where your own district stands, ask whoever manages its technology contracts whether the standard has been added. Don’t assume.

Your lesson plan doesn’t change on November 1. Nothing in the standard tells you which tools to use or how to use them. It doesn’t approve any product for any grade, and none of the official announcements I read lists the Microsoft products it covers. It sets rules for the vendor about data and safety.

Whether a given activity is worth doing with AI at all is still your call. The standard was never meant to make that call for you.

The tools it does not cover, which is most of them

This is one company. The standard binds Microsoft and nobody else. The image generator a colleague recommended, the quiz app with an AI button, the chatbot a student opens on a phone, the browser extension that summarizes a reading: none of them signed anything on September 9.

Two other AI companies say they are talking to the union. Education Week ran an Associated Press report quoting OpenAI, which said it looks forward to finalizing its own agreement, and Anthropic, which said it is working with the union on a standard. The same report says Google had not said whether it would offer similar protections. Until a deal is signed and published, a conversation is not a contract.

That matters most for the tools nobody formally chose. In my experience, AI often gets into a classroom through one teacher’s free account rather than through a district purchase. A free account falls under the company’s general terms, not anything your district negotiated.

There is a second route too. If you read our piece on Gemini being switched on in Google Classroom, you already know that AI also arrives inside platforms schools already pay for. In that case the protection is whatever that contract happens to say.

Three questions to ask before a class uses any AI tool

The Microsoft standard answers these three questions for Microsoft. Ask them about every other tool, and if nobody can answer, that tells you something too.

1. Does what students type train the company’s models?

Search the privacy policy for “train” and “improve our models,” and check whether education accounts are treated differently from personal ones. If the policy says nothing, I would assume the answer is yes.

2. Who decides when the data is deleted?

The standard gives the school control over retention and deletion. For any other tool, find out whether your district can ask for data to be deleted, or whether the company keeps it as long as it likes.

3. Is the promise in a contract, or only on a web page?

A privacy page can be rewritten any Tuesday. The Microsoft terms carry weight because a district can enforce them. Ask whether your district has a signed agreement with the vendor, or whether students are using the tool under the default terms anyone accepts with one click.

You don’t need a lawyer for any of this. You need ten minutes with the person in your district who handles technology. If the answers are vague, keep using the tool for your own preparation and keep student accounts out of it.

What I would not read into it

I would not read it as proof that AI in class is now safe. A privacy standard governs what happens to data. It says nothing about whether a generated text is accurate, whether a worksheet suits the level of the class, or whether a student learned anything. That still depends on the adult checking the output.

I would not read it as anonymity either. In its coverage of the agreement, Education Week notes that de-identifying data does not guarantee a person can never be identified, because AI can cross-reference information to point back to an individual. Using data in fewer ways is a real gain. It is not the same as no risk.

And I would not treat it as the end of the conversation in your building. The standard is a minimum for one vendor, and it gives you a template for questioning the rest. For a teacher, that may be its most useful side: a published list of commitments to point to when someone asks why you want to know what a tool does with what your students type.

Frequently asked questions

Does Microsoft use student data to train AI?

Under the National AI Safety & Privacy Standard announced on September 9, 2026, student and educator data will not be used to train AI models, sold or repurposed. Education Week reports a narrow exception for specific safety-related data. The protections apply when a district adds them to its Microsoft agreement.

When does the Microsoft AI privacy standard for schools take effect?

The protections become available to every US school district on November 1, 2026. Districts can add them to new or existing Microsoft agreements without renegotiating or renewing.

Does the standard cover Google, OpenAI or other AI tools in school?

No. It binds Microsoft only. OpenAI and Anthropic have said they are working with the AFT on agreements of their own, but as of mid-September 2026 neither had announced a signed one. Google had not said whether it would offer similar protections.