Photo via Mother Jones

A Teachers Union and Microsoft Just Made An AI Safety Deal. But Compliance Remains an Open Question.

Read the original at Mother Jones ↗
Cite

Nguyen, A. (2026, September 10). A Teachers Union and Microsoft Just Made An AI Safety Deal. But Compliance Remains an Open Question. Mother Jones. https://www.motherjones.com/politics/2026/09/teachers-union-microsoft-deal-nyc/

Dated when The Aporia collected the article, usually within hours of publication. Check the original for its exact date and byline.

Mother Jones · collected 2026-09-10 · by Alex Nguyen

Quick Summary

The American Federation of Teachers (AFT) announced an agreement with Microsoft on Wednesday that prohibits the company from tracking student use of its AI tools and using student and teacher data to train AI models. The guidelines apply to any school district that signs a customer agreement with Microsoft, giving them 180 days to delete records of AI data use upon request. Microsoft can still collect some user data for "narrow safety and security" situations, such as detecting harm or security threats. The AFT's President Randi Weingarten mentioned ongoing conversations with other tech companies, including OpenAI and Anthropic.
Written by AI on 2026-09-10, from this article's own text only (the story summary below draws on all the coverage).

AI analysis

Story summary

Microsoft has agreed to a deal with the American Federation of Teachers (AFT) that restricts the tech giant's ability to track student use of its AI tools and apply student and teacher data to train AI models. The agreement prohibits Microsoft from using student data for training, but allows exceptions in certain "narrow safety and security" situations, such as detecting or preventing harm. The guidelines also require Microsoft to delete records of AI data use from students and school districts on its active systems within 180 days of a request. This deal applies not only to the AFT, but also to any other school district that wants to follow these principles when purchasing Microsoft's AI products. The agreement was announced in a Manhattan press room, attended by about 50 reporters and union members.

Written for “AI Safety Agreement” on 2026-10-30, grounded in this article and the 0 other(s) covering the same event.

Story

📰 AI Safety Agreement
Technology · 1 article(s) covering the same event.

Compared with similar articles

This article reads leans left and hedges 5% of its claims. Each row says how that neighbour differs.
The Straits Times
⚖️ Leans left Hedges 1 of 14 claims Outlet record 66/100
“The articles describe different aspects of Microsoft's involvement with AI safety and education, not the same specific incident.”
Platformer
⚖️ Leans right further right than this Hedges 0 of 3 claims Outlet record 96/100
“Article A describes a press conference announcing an agreement between Microsoft and a teachers union, while Article B discusses broader themes related to AI regulation and industry trends.”
Semafor
Hedges 1 of 32 claims Outlet record 60/100
“Article A describes a press conference announcement between a teachers' union and Microsoft regarding AI safety, while Article B discusses broader themes about AI companies and regulation in Washington.”
Semafor
⚖️ Leans left Hedges 2 of 7 claims Outlet record 60/100
“The articles discuss different aspects of AI regulation and agreements, with Article A focusing on a deal between Microsoft and a teachers union, while Article B discusses lack of consensus in Congress about AI regulation.”
Toronto Star
⚖️ Leans strongly left further left than this Hedges 0 of 24 claims Outlet record 64/100
“The articles mention different topics, one about social media safety for children in Canadian schools and the other about an agreement between a US teachers' union and Microsoft regarding AI tool usage.”
The Straits Times
Hedges 3 of 24 claims Outlet record 66/100
“The articles cover different events: one is about an AI researcher quitting Anthropic, while the other is about a deal between a teachers' union and Microsoft regarding AI safety guidelines”
The Hindu
Hedges 1 of 51 claims Outlet record 60/100
“Article A describes a policy announcement by New York's Mayor and Schools Chancellor, while Article B reports on an agreement between a teachers' union and Microsoft, which does not mention the moratorium announced in Article A”
Washington Examiner
⚖️ Leans left Hedges 4 of 60 claims Outlet record 70/100
“Article A discusses a general debate about AI's impact on jobs, while Article B reports on a specific agreement between Microsoft and a teachers' union, which is not mentioned in Article A”
The Intercept
Hedges 7 of 77 claims Outlet record 97/100
“Article A discusses a request from the Pentagon to OpenAI for a custom AI tool, while Article B reports on an agreement between a teachers union and Microsoft with guidelines for AI use in schools”
The AI safety vibe shift Different event
Platformer
Hedges 10 of 67 claims Outlet record 96/100
“Article B discusses AI safety warnings and industry concerns, while Article A reports on an agreement between Microsoft and a teachers' union regarding AI tool usage and data privacy”
How this article was measured signals, leaning charts, publisher record, byline, topics and claims

Signals How these are calculated →

Claims extracted
21
sentences that state something as happening or true
Uncertain
5%
1 of 21 hedged ("reportedly", "may", "according to")
Leaning
Leans left
of the writing, not the subject · beta estimate
Publisher signals
95/100
Mother Jones's corrections and hedging, across everything collected from it. Not a measure of accuracy.
Outlets on this story
1
Technology
Same framing elsewhere
1
articles that frame the event the way this one does
Analyzed 2026-09-10 · how these are computed

How this is being covered How these are calculated →

Article leaning vs. publisher reliability
Source leaning vs. consistency

Publisher

Mother Jones · 406 article(s) · 0 correction(s) detected
No corrections detected for this publisher. That may mean careful reporting, or simply that nothing has been checked.

Who wrote this

Alex Nguyen
43 article(s) here · 1 carrying a prediction
🔮 “Never before in American history has the federal government, on the eve of an election, sought to coerce a state and local government to abandon their own Constitutional elections procedures and bend to the will of the Executive Branch,” Judge David Carter wrote in a 26-page ruling.
2026-10-29 · assertive framing · Court Denies White House Bid to Toss California Ballots
🔮 There was Serena Williams, Charlize Theron, Matthew McConaughey, Jimmy Kimmel, and even a closing set from my fave Charli XCX , including her song “Apple,” which she introduced to the crowd as a song “about generational trauma.”
2026-10-29 · assertive framing · Apparently There Is an Airbnb Gala
🔮 The issue may prove to be a deciding factor in many races, as Democrats would need to flip four seats each in both the US House and Senate to build a simple majority.
🔮 The AI giant Anthropic disclosed its initial public offering filing with the SEC on Monday in what is expected to become one of the largest stock debuts in history.
🔮 President Donald Trump ramped up his Islamophobic remarks against Democratic Senate candidate Abdul El-Sayed at a Sunday campaign rally in Michigan, hurling everything from unfounded “jihadist” accusations to suggestions that an El-Sayed victory would mean canceling Christmas.
🔮 The problem is reaching a level where banks have been forced to scale up their investment of resources into preparing for breaches and restoring customer access as quickly as possible—which, given other recent developments in AI-powered cybersecurity breaches, might not have been a bad idea anyway.
🔮 In the five-minute confrontation that has now gone viral, the young man, who the Handbasket first identified as Thomas “Tommy” Teepell, asked El-Sayed if he would “disavow” the Muslim Brotherhood.
🔮 Elon Musk’s SpaceX and AI giant Anthropic have said they will give billions of dollars in shares to the public as general sentiment around the technology’s economic and societal consequences continues to plummet.
2026-10-20 · assertive framing · One SpaceX Share Per Child: The New AI PR Strategy
🔮 Construction to create “one of the Best and Most Spectacular Golf Courses anywhere in the World” at East Potomac Park would start on November 15, the president wrote on Truth Social.
🔮 The five-minute conversation took place in Caro, Michigan, where a man who identified himself as “Tommy” asked from the audience: “Will you disavow the Muslim Brotherhood?”
Also by Alex Nguyen
Apparently There Is an Airbnb Gala
2026-10-29 · Mother Jones
Nothing else under this byline is closely related to this article, so these are simply their most recent.
All 43 articles by Alex Nguyen →

Topics

AFT Manhattan Microsoft OpenAI the American Federation of Teachers’

Subjects

Microsoft ORG · 12× AFT ORG · 3× New York City GPE · 2× Smith PERSON · 2× Anthropic ORG · 1× Brad Smith PERSON · 1× Manhattan GPE · 1× OpenAI ORG · 1× Randi Weingarten PERSON · 1× the American Federation of Teachers’ ORG · 1×

Narrative

When I asked Microsoft Vice Chair and President Brad Smith about specific guidelines on how an accountability process will work if Microsoft fails to comply with remediation on Wednesday, he largely dodged the question, saying that Microsoft has to publish reports that provide transparency to the AFT and school districts and have tight deadlines for fixes.
framing: assertive · carried by 1 article(s) · first seen 2026-09-10
🔮 The current guidelines do not present explicit, quantifiable rules for what is considered remediation by a “reasonable” time—it’s something that both Microsoft and the customer will “engage in good faith” to come up with.

Claims (21 extracted, 1 hedged)

On Wednesday morning, in a Manhattan press room containing about 50 reporters and union members, one of the largest teachers’ unions in the US announced an agreement with Microsoft that, among other things, prohibits the tech company from tracking student use of its AI tools and applying student and teacher data to train AI models. asserted
that → contain → models
The guidelines listed in the American Federation of Teachers’ deal with Microsoft apply to any school district “that desires the same principles” when forming their own customer agreements with the company and purchasing its AI products. asserted
that → list → products
The union also gives Microsoft—and any other tech company that signs on (AFT President Randi Weingarten mentioned ongoing conversations with OpenAI and Anthropic during Wednesday’s announcement)—a 180-day deadline to delete records of AI data use from students and school districts on its active systems following a request. asserted
Weingarten → give → request
Microsoft agreed not to use student and teacher data to train its AI — but there are exceptions. asserted
Microsoft → agree → AI
The guidelines point out “narrow safety and security” situations, including letting the company use “the minimum amount” of school user data to detect, prevent, investigate, or remediate harm to users or security threats to its AI products and retain it “for as long as reasonably necessary.” asserted
company → point → it
The contract also lists “contractual remedies” for school districts that sign on, such as the ability to pause or discontinue Microsoft’s deployment of its AI systems if the company commits any material breaches of its agreement obligations. asserted
company → list → obligations
The current guidelines do not present explicit, quantifiable rules for what is considered remediation by a “reasonable” time—it’s something that both Microsoft and the customer will “engage in good faith” to come up with. asserted
Microsoft → present → faith
When I asked Microsoft Vice Chair and President Brad Smith about specific guidelines on how an accountability process will work if Microsoft fails to comply with remediation on Wednesday, he largely dodged the question, saying that Microsoft has to publish reports that provide transparency to the AFT and school districts and have tight deadlines for fixes. asserted
that → ask → fixes
“Obviously, this [remediation process] becomes part of our broader contract with each school district, and all of those [existing] contracts already have built-in remedies,” Smith told Mother Jones. asserted
Smith → become → Jones
“I think that there’s just an unprecedented amount of transparency associated with the commitments that are in this agreement.” asserted
that → think → agreement
When I asked again about specific steps that will be taken if Microsoft fails to comply, Smith said they are “connected back into the provisions that we have in our contract with each school district. asserted
we → ask → district
Those can vary.” asserted
Those → vary → ?
“Mostly, what I find is that what people want when something goes wrong is just to go fix it. asserted
something → find → it
People don’t want to end up with more time with lawyers,” he continued. asserted
he → want → lawyers
“That’s always an option if somebody wants to take action, but let’s just fix problems when they arise.” asserted
they → ’ → problems
While accountability applies to how each school district will work with Microsoft in the near future, the company seemingly not holding discussions now raises questions about the feasibility of contract enforceability. asserted
company → apply → enforceability
Microsoft, and other companies like Google, have a history of promoting their AI products to schools and educators, and a healthy skepticism is needed when relying in part on tech companies to police AI and previous failures of implementing educational technology. asserted
skepticism → have → technology
Weingarten and Michael Mulgrew, president of the United Federation of Teachers representing New York City public school educators and others, who were also taking questions immediately following the announcement, did not respond to the question. asserted
who → represent → question
The contract comes a week after New York City Mayor Zohran Mamdani announced a one-year moratorium on AI use for elementary and middle school students in the city’s public schools, and the Los Angeles Unified School District announced restrictions to generative AI platforms for all students for the 2026-27 school year. asserted
District → come → year
The AFT’s contract with Microsoft contains the sentence “When in doubt, protect the student.” on the bottom of each of the 31 pages. asserted
contract → contain → pages
But exactly how any of these policies around AI use in schools will actually work remains unclear. uncertain
work → work → schools
💬Give feedback
🕘History 🎫Support