The Frank
Home
Newsletter
About
Subscribe
Microsoft Bars Training AI on Student Data

Microsoft Bars Training AI on Student Data

The Frank Staff

The Frank Staff

The Frank Staff.
news@thefrank.com
@TheFrank_com

Sep 9, 2026

·

0 min read

Share options

Email
Facebook
X
Telegram
WhatsApp
Reddit
Randi Weingarten, American Federation of Teachers president, speaking into a microphone
Randi Weingarten, AFT president, at the podium unveiling the AI privacy standard with Microsoft for schools.Illustration: The Frank

The News

Microsoft agreed Wednesday not to train its AI models on student or teacher data, under a new privacy standard written with the country's second-largest teachers union that school districts can enforce through their contracts.

Microsoft and the American Federation of Teachers announced a "National AI Safety & Privacy Standard" on Wednesday in Manhattan — a set of 10 protections districts can write directly into their Microsoft agreements and enforce as contract terms.

The standard bars using student or educator data to train AI models, bars tracking student use of the tools, bans AI companion chatbots for students, and requires a human to review "high-risk" decisions. Districts that adopt it can pause or cancel Microsoft's AI deployment and seek damages over a material breach.

It does not ban AI in classrooms. It comes a week after the nation's largest school system, New York City, imposed a one-year moratorium on AI for elementary and middle school students, and Los Angeles Unified restricted generative AI for all students this school year.

The 10 Protections

Along with the training and tracking bans, the standard prohibits selling student data or using it for advertising, limits how much data Microsoft collects in the first place, requires plain-language disclosure to families of how the tools work, sets a 72-hour breach reporting requirement, and subjects Microsoft to annual certification and audit rights. Districts can demand deletion of AI data records from active systems within 180 days.

There are carve-outs. The guidelines allow Microsoft to use "the minimum amount" of school user data in "narrow safety and security" situations — detecting, preventing, investigating or remediating harm or security threats — and to keep it "for as long as reasonably necessary," Mother Jones reported.

Timeline

2024

The Kids Online Safety Act, which would impose a broader duty of care on platforms used by minors, passes the Senate. It has stalled since being reintroduced last year, Fortune reported.

April 2026

More than 250 child-safety experts and groups call for a moratorium on AI in schools, citing student data risks.

May 2026

AFT President Randi Weingarten calls for a screen ban before third grade, a ban on student-facing AI before middle school, and a ban on companion chatbots for students under 16.

Sept. 2, 2026

New York City Mayor Zohran Mamdani and Gov. Kathy Hochul announce a one-year AI moratorium for the city's elementary and middle school students; Los Angeles Unified announces generative AI restrictions for all students for 2026-27.

Sept. 9, 2026

Microsoft Vice Chair and President Brad Smith announces the standard alongside Weingarten and United Federation of Teachers President Michael Mulgrew.

Nov. 1, 2026

The protections take effect for every school district Microsoft works with, whether or not the district acts, Smith said. Districts can also fold the terms into new or existing agreements without renegotiating the whole contract.

Reactions

Weingarten said the union acted because Washington did not. "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," she said. She added: "We can get angrier and angrier, or we can act decisively; anything less than legally enforceable provisions is simply a wish list."

She called existing law inadequate — "HIPAA and FERPA never envisioned the advent of AI" — and said the standard is "a very important first step" toward safeguards in law.

Smith described the document in three words. "It is 30 pages, but in some ways it comes down to three words: privacy, safety, and transparency," he said. "I actually think the best dose of medicine, if you will, for anybody is sunshine transparency."

Mulgrew, who represents roughly 200,000 members in New York City, said the deal "starts to level the playing field in a competitive, ever-changing arena that lacks the guardrails our children and school communities need." He defended the city's moratorium: "We have been clear that we do not believe [AI use in] K through 8 is appropriate at this point in time."

Enforcement is where reporters pushed back. Asked by Mother Jones what specifically happens if Microsoft misses a remediation deadline, Smith pointed to existing contracts and transparency reports rather than a set process: "Obviously, this becomes part of our broader contract with each school district, and all of those contracts already have built-in remedies." Pressed again, he said the remedies "can vary" and added, "Mostly, what I find is that what people want when something goes wrong is just to go fix it." The guidelines do not define a "reasonable" remediation time, leaving Microsoft and the district to "engage in good faith."

What's Next

Weingarten said OpenAI and Anthropic — Microsoft's partners in the union-backed National Academy for AI Instruction — have "both expressed willingness to do this kind of agreement, and I am hopeful that they will sign soon." Neither company responded to Fortune's requests for comment.

The terms become available to districts in November, and Microsoft faces its first annual certification and audit obligations under the standard after that. New York City's and Los Angeles' one-year restrictions run through the 2026-27 school year while both districts weigh permanent rules.

More

Federal law has not caught up. The Children's Online Privacy Protection Act requires parental consent before companies collect data from children under 13, but it predates generative AI by more than two decades and says nothing about how AI models are trained.

The AFT is not a pure adversary of the industry here: the union announced a $23 million AI training hub for educators funded by Microsoft, OpenAI and Anthropic. Microsoft and Google have both marketed AI products aggressively to schools.

The AFT document carries one line at the bottom of every page, according to Mother Jones: "When in doubt, protect the student."

Poll

Is a contract with Microsoft enough to protect student data?

Yes — enforceable terms beat waiting on Congress
0.0%
No — only a federal law will do
0.0%
No — districts can't police a tech giant
0.0%
Not sure
0.0%
0 votes
·
Final result

Share options

Email
Facebook
X
Telegram
WhatsApp
Reddit

Charleston White Arrested Again on Florida Battery Charge

Sep 13, 2026

Border Patrol Rescues MS NOW Crew in Desert

Sep 13, 2026

Supreme Court Restores Missouri's Old House Map

Sep 13, 2026

Alaska Drops Felony Voter Charges Against Samoans

Sep 13, 2026

Trespasser Breaches Harris' Malibu Estate; No Arrest

Sep 12, 2026

Hegseth Pressed Downed Airmen Into 60 Minutes

Sep 12, 2026

Houthis Seize Mokha, Choking Saudi Oil Exports

Sep 12, 2026

Altman Rules Out OpenAI IPO This Year

Sep 12, 2026

Jury Acquits Lil Durk in Murder-for-Hire Case

Sep 12, 2026

Judge: Trump Broke Law Halving FEMA Staff

Sep 12, 2026

Maher Calls Democrats' Gaza Genocide Claim a Lie

Sep 12, 2026

Blizzard Revives StarCraft, Sends Diablo to Netflix

Sep 12, 2026

Trump's Queens Childhood Home Sells for $1.93M

Sep 12, 2026

Scientists Find New Virus in Tick Invading US

Sep 12, 2026

Britain Bans Settlement Imports; US States Threaten Payback

Sep 12, 2026

Prison Fire Kills 7 as Kurds Protest Killing

Sep 12, 2026

Kosovo Rallies for Ex-President Before Hague Verdict

Sep 12, 2026

Cruz Booed at GameDay Over College Sports Bill

Sep 12, 2026

Blast Hits Bulgarian Arms Plant Supplying Ukraine

Sep 12, 2026

DOJ Loses Nearly Every Protester Assault Trial

Sep 12, 2026

  • About
  • Contact
  • Policy & Terms
  • Recaptcha