Education news

Schools Get an AI Privacy Deal, Not Proof It Teaches

Teachers' unions and Microsoft signed schools an iron-clad AI privacy standard, and the protections are real. But read the whole thing and you will not find one line about whether these AI tools actually teach a child anything. That gap is the story.

Laura Lurns Last updated September 13, 2026 · 10 min read

Schools Get an AI Privacy Deal, Not Proof It Teaches

If your child’s school uses an AI tool, you have probably wondered two separate things about it. Is my kid’s data safe? And is this thing actually helping them learn? On September 9, the country’s largest teachers’ unions and Microsoft answered the first question with a binding national standard, and left the second one almost entirely untouched. The privacy protections in that deal are real and hard-won. What is missing from it is any claim that the products it covers teach a child anything at all, and that absence is the whole story.

TL;DR

On September 9, 2026, the AFT, the UFT and Microsoft announced a binding National AI Safety and Privacy Standard for U.S. schools, available to every district on November 1.
Its protections cover data use, student tracking, human oversight, transparency, breach notification and a district's right to end agreements and pursue damages, and the AFT says OpenAI and Anthropic are in talks to reach similar terms.
The standard makes no claim about instructional effectiveness; Fairplay's Josh Golin said on the record there is no evidence AI products are effective teaching tools for children.
The learning science names retrieval practice, spaced practice, step-level tutoring and task-focused feedback as what makes educational technology teach, and those studies were run on earlier systems, not on the generative-AI tools now in classrooms.
A week earlier, on September 2, New York City placed a one-year moratorium on student-facing generative AI for grades 2-K through 8, covering roughly 600,000 students; the privacy standard does not lift or address it.

Context the coverage leaves out

What the coverage gets wrong

Most of the reporting frames the September 9 standard as a straightforward win for students, and read as a privacy story it earns that. The trouble is what the framing implies but never says: that because these AI tools are now safer and more private, they are good for classrooms. That step does not follow, and no party to the deal claims it. The agreement is about data handling, human oversight and contract enforcement, from end to end, and it makes no representation about whether the covered products actually teach. The one voice in the coverage naming that gap is a critic, Fairplay's Josh Golin, and his point is the load-bearing one: privacy and teaching effectiveness are different questions, and answering the first does not answer the second. A parent reading only the headline would come away thinking the hard question had been settled. It was not addressed.

The new AI standard has parents asking what it actually protects, and how to tell whether the AI tools their kids use are teaching them anything. Here are the questions coming up most.

Common questions

What does the AFT, UFT and Microsoft AI standard actually cover?

Data and accountability, not teaching. Microsoft agrees not to use student or teacher data to train its AI models except in limited safety cases, not to track students, and not to let its AI make decisions without human oversight. Districts get plain-language transparency, a 72-hour breach notice, and the right to end agreements and pursue damages for violations. It becomes available to every U.S. district on November 1, 2026. Nothing in it addresses whether the covered products help a child learn.

Does this mean AI tools are proven to help kids learn?

No. The standard is a privacy and safety agreement, and effectiveness is a separate question it does not touch. The learning science is clear about what makes educational technology teach, retrieval practice, spaced practice, tutoring that responds to a child’s specific error, and task-focused feedback, but those studies were run on earlier systems, not on the generative-AI tools arriving now. Whether today’s classroom chatbots teach that way has not been tested.

How do I tell if an AI learning tool is actually teaching my child?

Watch whether it makes your child do the hard part. A tool that asks them to recall an answer, try the next step, and stay with a problem is building a skill. A tool that hands over the answer, piles on points and streaks, and measures success in minutes-on-app is measuring engagement, not learning. A busy, happy child looks like a learning child, which is exactly why the two are easy to confuse.

My child is struggling and I do not know where to begin. What is a good first step?

Start with what you see at home, which no app dashboard shows you. A parent screener asks about what you are noticing across reading, writing, math and attention, and points you to where to focus today. A screener is a starting point, not a diagnosis. If your child might need formal accommodations such as an IEP or 504 plan, or you suspect a vision, hearing or medical cause, pursue a professional evaluation too, since that is the only route to those supports.

Share the idea

Teachers' unions and Microsoft won schools an iron-clad AI privacy deal. Real protection. But not one line says whether these AI tools actually teach a child anything. That question is still yours.

Post on X Follow

What happened

On September 9, 2026, the American Federation of Teachers, the United Federation of Teachers and Microsoft jointly announced a National AI Safety & Privacy Standard for U.S. schools, structured as a binding Memorandum of Agreement with open slots for other AI providers to sign on. AFT President Randi Weingarten and Microsoft Vice Chair and President Brad Smith signed it on September 7. “We have forged a hard-fought, iron-clad privacy agreement with real teeth that protects students and families, because no one else, including the federal government, has stepped up to do the real work,” Weingarten said.

The protections are substantive. Microsoft will not use student or teacher data to train or improve its AI models except in limited safety cases, is barred from tracking students, and is prohibited from letting its AI systems make decisions without human oversight. Districts get plain-language transparency about what data is collected, a 72-hour breach notification, and the power to end an agreement and pursue damages when a rule is broken. “This standard sets a high bar for child privacy and AI safety, and we’ll extend this agreement to every school district across the country,” Smith said. The protections become available to every U.S. district on November 1, and the AFT says it is in talks with OpenAI and Anthropic to reach similar terms.

The timing is pointed. A week earlier, on September 2, New York City Mayor Zohran Mamdani and Schools Chancellor Kamar H. Samuels put a one-year moratorium on student-facing generative AI for grades 2-K through 8, covering roughly 600,000 students, and disabled generative-AI features in more than 38 previously approved programs. “Children need teachers and human connection in order to learn and grow,” Mamdani said. The privacy standard does not lift that moratorium, and it does not speak to it. The two actions answer different questions.

There is no evidence that AI products are effective teaching tools for children and growing evidence that they undermine learning and create a number of risks for kids.
Laura Lurns · Learning Success expert

The frame the science supports

Here is the slide the coverage keeps making. A product that guards your child’s data, keeps a human in the loop, and holds a company liable sounds like a good product for a classroom. Safe and private start to read as good for learning. They are not the same claim, and nobody in this deal made the second one. Josh Golin, who runs the children’s advocacy group Fairplay, said it plainly the same day: “There is no evidence that AI products are effective teaching tools for children and growing evidence that they undermine learning.” His group is calling for a pause on student-facing generative AI in schools until the effectiveness question is actually answered.

Start with what is not in dispute, because good ed-tech is real and the honest line is narrow. The apps and programs that teach are the ones built on things the learning science nailed down long ago. Roediger and Karpicke showed that being tested is itself the learning, not merely the measurement of it, which is why software that makes a child retrieve an answer beats software that lets them reread. Cepeda and colleagues showed that practice spread out over time sticks far better than the same practice crammed together. VanLehn found that tutoring systems which respond to a child’s specific error, step by step, approach the power of a human tutor, while software that only checks the final answer barely moves the needle. Kulik and Fletcher confirmed the step-level version outperforms ordinary instruction, with the sharp caveat that gains look bigger on tests built to match the program than on independent ones. And the feedback research, from Kluger and DeNisi through Hattie’s group, is blunt that feedback is powerful but not automatically good: a next-step nudge helps, while a red X, a streak, or a leaderboard often backfires.

Notice what the September 9 standard names, and what it does not. It names data, oversight, transparency and enforcement. It does not name retrieval, spacing, step-level response or task-focused feedback, because it was never about how the products teach. And the honest limit runs deeper than that: the tutoring research above was done on earlier intelligent-tutoring systems, not on the generative-AI chatbots now arriving in classrooms. Whether a GPT-style assistant behaves like a step-level tutor that catches the specific mistake, or like answer-only drill dressed up in conversation, has not been tested. That is an open question, and a privacy agreement does not close it. This is not an argument against technology. It is an argument for asking the question the science says decides everything: does it teach the way learning actually works, or does it just keep a child busy?

Key takeaways

The gap: The standard governs student data and oversight, and says nothing about whether the AI tools teach.
What teaching needs: The science says retrieval, spacing, step-level response and task-focused feedback are what make ed-tech work.
Untested ground: Whether a classroom chatbot tutors step by step or just drills answers has not been studied.

What it means for your child

Take the win first, because it is genuine. If your district signs on, your child’s schoolwork will not be quietly feeding a company’s next model, a human has to stand behind consequential decisions, and the district has a lever to walk away when a vendor breaks the rules. A year ago none of that was guaranteed. Golin’s own caution is worth carrying alongside it: the standard covers Microsoft and not the many other companies selling to schools, and it puts the work of enforcement on districts, which is a heavy thing to hand a school office. So it is progress with edges, not a finished job.

What the deal does not decide for you is whether any given tool actually helps your child learn, and that judgment is better made at your kitchen table than left to a contract. There is a simple test. Watch whether the tool makes your child do the hard part or does it for them. A program that asks them to recall, to try the next step, to sit with a problem long enough to work it out is building something. A program that hands over the answer, showers points and streaks, and measures success in minutes-on-app is measuring engagement, not learning, and the two are easy to confuse precisely because a busy, happy child looks like a learning one. If you want to see what the process looks like when it is built around how learning works, our at-home approach starts by asking what you are already seeing in your child, then names which skill to build first.

Treat this the way a thoughtful parent should treat California’s push for mandatory dyslexia screening: a real step that helps or falls short depending entirely on how it is used, not a reason to hand the decision off. The privacy standard buys accountability. The New York moratorium buys a pause. Neither one answers whether a product teaches, and neither one is meant to. That question stays with the adult who watches this child learn every day, and it is a better question than any of the ones the headlines are fighting over.

This standard sets a high bar for child privacy and AI safety, and we’ll extend this agreement to every school district across the country.
Laura Lurns · Learning Success expert

You already know the difference between a child building something real and a child being kept busy, because you watch it happen at your own table. That is the judgment this whole debate keeps trying to move somewhere else, into a contract, a commission, a company’s terms of service. The villain here is not the unions who fought for these protections, not Microsoft, and not the mayor who hit pause, all of whom are wrestling with a genuinely hard problem. It is the quiet assumption underneath the coverage: that a product which is safe, private and engaging must therefore be teaching. Engagement is not learning, and a child can move through a polished, well-guarded app for months while the underlying skills sit still. A points system makes a child feel like they are learning; real mastery makes them feel capable. Learning Success was built to refuse the substitution, on retrieval, effort and the relationship with the adult who loves the child rather than on streaks and badges. If you want a place to start building the real thing, our All-Access membership opens an assessment that asks about the processing systems your child’s learning runs on, then a roadmap that names which skill to build first.

See what All Access gives your child

Is your child struggling in school?

Get your free personalized learning roadmap

You describe what you see at home. We turn it into a plan you start this week.

Answer 5 short questionnaires about what you already notice, 30–45 minutes at your own kitchen table
Your child sits no test and gets no score: nothing to schedule, nothing for them to dread
You do the answering, the AI does the writing, and a person reviews it before it reaches you
Access all 40+ courses instantly: reading, math, focus, processing and more, with new ones added regularly

Why we use AI, plainly: it writes from a knowledge base our team maintains and audits. We work through it line by line and pull anything the evidence stops supporting. The roadmap you get on Tuesday reflects what we corrected on Monday, and a human still reads it before you do.

Your school district must evaluate your child free of charge if you ask in writing, whatever your income and whatever the outcome (US, 34 CFR 300.111 and 300.301(b)). That route takes time and answers a different question than you do. This one starts today, from what you already know.

Get my free assessment & 7-day trial Cancel anytime during your trial.

Your answers stay yours. We do not sell your personal information, and we do not hand identifiable assessment data to outside AI companies to train their models.

A screener is a starting point, not a diagnosis. If your child might need formal accommodations (an IEP or 504 plan), or you suspect a vision, hearing or medical cause, pursue a professional evaluation too. That is the only route to those supports.

References

ABC7 New York — AFT, UFT, Microsoft set major AI schools announcement amid NYC classroom restrictions
Microsoft Source — AFT, UFT and Microsoft announce National AI Safety & Privacy Standard for schools
American Federation of Teachers — AFT, UFT and Microsoft announce National AI Safety & Privacy Standard for schools to protect students
Fairplay — Statement on AFT-UFT deal with Microsoft
Government Technology — Microsoft, AFT, UFT set precedent for district AI governance
Office of the Mayor of New York City — Mayor Mamdani and Chancellor Samuels put students first

Laura Lurns · Learning Success expert Writes about the learning brain for parents who want plain answers. Every article is grounded in current neuroscience and classroom practice.