The AI Safety Standard for Schools: What It Actually Binds

The new AI safety standard schools are being offered this autumn is not a law, not a state rule, and not something your district has to follow. It is a contract clause. On September 9, 2026, the American Federation of Teachers, the United Federation of Teachers and Microsoft announced a “National AI Safety & Privacy Standard” for schools, and the mechanism is the whole story: districts can write the protections directly into their Microsoft customer agreements, which makes them enforceable with termination rights and damages. Nothing obliges a district to do it. That distinction decides whether any of this reaches your child.

Key Takeaways

• The standard binds Microsoft, not your school, and only once your district amends its Microsoft agreement. There is no penalty for a district that skips it.
• It covers three things: no training AI models on student or educator data, human review before AI-influenced decisions, and plain-language disclosure to families.
• It applies to Microsoft products only. Google Workspace, OpenAI, Khan Academy, Canva, Securly and every other vendor in your child’s school are untouched by it.
• Federal law already sets a floor — FERPA for education records, COPPA for children under 13. Neither one lets a parent sue. Both are enforced by agencies.
• The one question worth asking your district office: did you amend the Microsoft agreement, and on what date?

What the September 9 announcement actually said

A vendor standard is a set of promises a company makes about its own product, written in language a customer can hold it to. That is what this is.

The AFT’s announcement, dated Wednesday, September 9, 2026, sets out three commitments. On privacy: student and educator data cannot be used to train AI models, cannot be sold, and cannot be repurposed; the school keeps control over how data is used, how long it is retained, and when it is deleted. On safety: AI systems must include security controls, safeguards and human oversight, AI cannot make decisions without human review, and the systems must avoid “harmful or manipulative experiences.” On transparency: families get clear information about where AI is used and what data is collected, and schools keep authority over when and how AI is deployed at all.

AFT president Randi Weingarten called it “a hard-fought, iron-clad privacy agreement with real teeth that protects students and families.” Microsoft’s Brad Smith said the standard “sets a high bar for child privacy and AI safety, and we’ll extend this agreement to every school district.” UFT’s Michael Mulgrew framed it as giving families “the information and privacy protections they have been demanding.”

The teeth are real but specific. U.S. districts can incorporate the protections into their Microsoft customer agreements. Once incorporated, a district can end the agreement and seek damages if Microsoft violates them. That is a commercial remedy held by the district, exercised by district lawyers.

One date discrepancy is worth naming. The AI-in-education policy tracker that logged the announcement on September 9, 2026 reports that districts can begin writing the protections into Microsoft contracts starting November 1. AFT’s own release describes the contractual mechanism but does not name November 1. If your district tells you the standard is “in effect,” ask which document and which date they mean.

Who this binds, and who it leaves out

Here is the chain, stated plainly. Microsoft makes the promise. The AFT and UFT negotiated it. A district accepts it by amending a purchasing contract. A parent is not a party to that contract at any point.

That is not a criticism of the standard. It is how vendor agreements work, and it matters because the enforcement path runs through the district, not through you. If a district never amends its agreement, the standard changes nothing in that district. If it does amend, a parent who believes the terms were breached would raise it with the district, and the district decides whether to act.

The scope limit is sharper. Your child’s school almost certainly runs software from a dozen companies. A standard negotiated with Microsoft reaches Microsoft 365, Teams, Copilot and the rest of that stack. It does not reach a reading app the third-grade team picked, a chatbot a high school teacher discovered on her own, or a monitoring tool the IT department bought in 2023.

Instrument	Who it binds	Who enforces it	Covers AI training on student data?	Can a parent act directly?
Microsoft–AFT AI standard (Sept 2026)	Microsoft, once a district signs	The district, via contract	Yes, prohibited	No
FERPA (34 CFR Part 99)	Schools receiving federal funds	U.S. Dept. of Education	Not addressed	No private suit; file a complaint
COPPA (FTC Rule)	Operators collecting data from under-13s	Federal Trade Commission	Indirectly, via consent rules	No; report to the FTC
State AI rules (e.g. Florida, Sept 2026)	School districts in that state	State education agency	Varies by state	Through state process
Your district’s own AI policy	District staff and students	The superintendent	Whatever it says	Through the board

How the federal floor already works

Two federal laws cover most of what parents assume a school AI policy covers, and both are older than any chatbot.

FERPA, the Family Educational Rights and Privacy Act, is described by the U.S. Department of Education’s Student Privacy Policy Office as “a federal law that affords parents the right to have access to their children’s education records” along with the ability to request amendments and control over disclosure of personally identifiable information. The regulations sit at 34 CFR Part 99. One detail parents of older teens miss: when a student turns 18 or enters a postsecondary institution at any age, those rights transfer from the parent to the student. A high school senior who is already 18 holds her own FERPA rights.

COPPA, the Children’s Online Privacy Protection Rule, applies to “operators of websites or online services directed to children under 13 years of age,” and to operators with actual knowledge they are collecting personal information from a child under 13. The current rule text was published in the Federal Register on April 22, 2025. Note the age cutoff. A 14-year-old using an AI tool for homework is outside COPPA entirely.

Neither statute gives you a courtroom. FERPA complaints go to the Department of Education; COPPA complaints go to the FTC. That is precisely the gap the Microsoft standard is trying to fill with contract law instead, and it is a reasonable thing to try. It is also why “iron-clad” deserves a footnote: the clad iron belongs to whoever signed.

Why the “no training on student data” clause is the important one

Of the three commitments, the training prohibition is the one that changes engineering, not paperwork.

When a model is fine-tuned on text, some of that text can be reconstructed from the model later under the right prompting conditions. That is the mechanism behind every “the AI quoted my document back at me” story. A promise not to train on student work is therefore a promise about the model’s weights, not just about a database. It means the essays your eighth grader writes in Word cannot become part of the statistical substrate of a future model release.

The second clause, human review before AI-influenced decisions, is doing quieter work. Schools make consequential decisions about children constantly: placement, intervention tiers, discipline referrals, special-education eligibility. A requirement that no decision issues from a model without a human in the loop is the difference between a tool that drafts a recommendation and a system that assigns a label. Ask your district which of its decisions currently touch an AI-generated score.

The transparency clause is the weakest of the three, for a structural reason. “Clear information about AI usage” is satisfied by a notice. Most districts already publish notices. If your district’s current AI notice is a paragraph on page four of the handbook, the standard does not by itself make it better.

What to actually do with this at home

Ask the one contract question

Email the district’s technology director or data privacy officer with a single question: “Has the district incorporated the September 2026 AFT–Microsoft AI safety and privacy terms into its Microsoft agreement, and if so on what date?” You will get one of three answers: yes with a date, no, or confusion. All three are informative. A district that cannot answer has not read the standard.

Get the vendor list, not the policy

The policy tells you what the district intends. The vendor list tells you what is running. Most districts maintain an approved software inventory, often published for state reporting. Ask for it, then look for the names that are not Microsoft. That is your real exposure map. Our guide to reading a school district AI policy in 20 minutes walks through what to look for once you have the document.

Separate the under-13 question from the over-13 question

If your child is 12, COPPA applies and the consent machinery is at least theoretically on your side. If your child is 14, it does not. Families with kids on both sides of that line should expect different answers from the same school about the same tool, and should not assume the older child is better protected. The opposite is usually true.

Check what AI is deciding, not just what AI is reading

Privacy questions get the attention. Decision questions matter more for a specific child. Ask whether any AI-generated output appears in a file that follows your child across grades: a reading-level estimate, a behaviour flag, a risk score. Those stick.

What not to do

Do not treat this standard as a reason to relax. It is a ceiling a district may choose to adopt for one vendor, and it says nothing about the three or four tools your child uses most. Parents who read a headline like this one and conclude the issue is handled are exactly the audience the headline was written for. The useful response is a question to your district, not reassurance.

What to Watch For Over the Next 3 Months

• Week 4: Board agendas. Contract amendments go through a vote in most districts. Search your board’s December and January agendas for “Microsoft,” “amendment” or “data privacy agreement.” If it is not there, it has not happened.
• Month 2 red flags: A district that announces it “follows the national AI standard” without naming a signed amendment. A vendor list that has grown since September without a corresponding privacy review. An AI notice in the handbook that still says “may use third-party tools.”
• Month 3 self-check: Can you name every AI tool your child used in school this term, and which company owns each one? If not, your map is the thing to fix, not the policy.

Frequently Asked Questions

Does this mean my child’s schoolwork is safe from AI training now?

Only in Microsoft products, and only if your district signed the amendment. The commitment is specific: student and educator data cannot be used to train models, sold or repurposed. It carries no weight over software from other companies, which is most of what a school runs.

Can I sue the school if the standard is violated?

No. The contract is between the district and Microsoft, so the district holds the remedies, including termination and damages. Parents raise concerns through the district or the board. FERPA and COPPA complaints go to the Department of Education and the FTC respectively.

Is the standard a federal regulation?

No. It is a private standard negotiated by two teachers’ unions with one company. State rules are a separate track entirely, and several states moved on their own schedule this autumn, including Florida’s parental opt-in requirement.

My district already has an AI policy. Does this replace it?

No. A district policy governs how staff and students may use AI. This standard governs what one vendor may do with data. A district needs both, and the two documents answer different questions. Our overview of what a school AI policy should say covers the district side.

What about AI chatbots my child talks to outside school?

Entirely outside this standard. Common Sense Media’s 2025 research found nearly three in four U.S. teens had used an AI companion, half of them regularly, and about a third had chosen an AI over a person for a serious conversation. That is a home conversation, not a contract clause.

How many kids are using AI for schoolwork anyway?

Pew Research Center found 26% of U.S. teens had used ChatGPT for schoolwork in its September–October 2024 survey of 1,391 teens, double the 13% recorded in 2023. The share rose with grade level, reaching 31% among 11th and 12th graders.

---

About the author

Ricky Flores is the founder of HiWave Makers and an electrical engineer with 15+ years of experience building consumer technology at Apple, Samsung, and Texas Instruments. He writes about how kids learn to build, think, and create in a tech-saturated world. Read more at hiwavemakers.com.

---

Sources

1. American Federation of Teachers. (2026, September 9). “AFT, UFT and Microsoft Announce ‘National AI Safety & Privacy Standard’ for Schools to Protect Students, Families and Educators.” AFT Press Center. https://www.aft.org/press
2. Pursuit. (2026). “AI in Education: News, Policies, Innovations” (entry dated September 9, 2026). https://www.pursuit.us/news/ai-in-education-news-policies-innovations
3. U.S. Department of Education, Student Privacy Policy Office. “What is FERPA?” Protecting Student Privacy. 34 CFR Part 99. https://studentprivacy.ed.gov/faq/what-ferpa
4. Federal Trade Commission. (2025, April 22). “Children’s Online Privacy Protection Rule (COPPA).” FTC Legal Library. https://www.ftc.gov/legal-library/browse/rules/childrens-online-privacy-protection-rule-coppa
5. Sidoti, O., Park, E., & Gottfried, J. (2025, January 15). “About a quarter of U.S. teens have used ChatGPT for schoolwork — double the share in 2023.” Pew Research Center. https://www.pewresearch.org/short-reads/2025/01/15/about-a-quarter-of-us-teens-have-used-chatgpt-for-schoolwork-double-the-share-in-2023/
6. Common Sense Media. (2025). “Talk, Trust, and Trade-Offs: How and Why Teens Use AI Companions.” https://www.commonsensemedia.org/research/talk-trust-and-trade-offs-how-and-why-teens-use-ai-companions