Microsoft and teachers unions launch AI contract standard for schools

Microsoft and two teachers unions on Wednesday unveiled what they describe as a first-of-its-kind national AI privacy and safety standard for schools — and, unlike most school AI frameworks so far, this one is designed to be inserted into vendor contracts. The Sept. 9 announcement from Microsoft, the American Federation of Teachers and the United Federation of Teachers says districts can add the protections directly to Microsoft customer agreements, turning familiar worries about student data, AI oversight and family notice into enforceable terms rather than voluntary promises.

The new standard would bar the use of student data to train AI, prohibit tracking students, require human oversight for school decisions, and obligate providers to give families and educators plain-language explanations of how AI tools work. But the protections are not automatic: the underlying memorandum says districts must ask for them, and providers can incorporate the terms through an existing privacy agreement, license agreement or addendum rather than by dropping in the text word for word. The document also says participating providers must make the substantive protections available to education customers on request within 90 days of the standard’s effective date. (aft.org)

That distinction matters. School systems have spent the last two years collecting AI principles, acceptable-use rules and pilot guidelines. What many district leaders still lacked was procurement language specific enough to survive contract negotiations. The Microsoft-AFT-UFT deal starts to fill that gap by tying high-level governance goals to a signed agreement and a district request process. In practice, that gives chief technology officers, legal teams and purchasing officers a template they can ask Microsoft to honor before a tool reaches classrooms at scale. (aft.org)

From guardrails to contract terms

The standard’s most important shift is not that it restates concerns schools already know well. It is that it translates them into procurement demands. An AFT fact sheet summarizing the agreement says the 10 principles include no training on student or educator data except for a narrow safety and security exception, data minimization, customer ownership of data, human oversight for high-risk decisions, clear family consent and transparency requirements, limits on feature creep and vendor lock-in, and long-term duties around deletion and breach response even after a contract ends. The same fact sheet says AI companions are prohibited. (aft.org)

The longer memorandum adds detail that school attorneys and privacy officers will notice. It says “covered data” includes prompts, responses, uploaded files, behavioral signals, metadata and derivatives, and that providers may not use that data to train, fine-tune, benchmark or otherwise improve any AI model outside the narrow safety and security exception. It also bars behavioral tracking, keystroke logging, continuous passive attention monitoring, long-term profiling and biometric collection such as facial recognition, voiceprints, gaze tracking and emotional inference. Those specifics go beyond the broad “trust us” language that often appears in ed-tech sales and privacy pages. (aft.org)

For district leaders, that specificity could be the real breakthrough. A superintendent or CTO who wants to say no to vague data practices now has a list of disallowed behaviors to take into negotiations. A teacher or principal who has worried that “AI oversight” is too fuzzy to mean much can now point to a standard that says AI may support decisions, but people decide. That does not eliminate implementation fights, but it does move them to a place schools understand well: contract review, approval workflows and renewal cycles. (aft.org)

A response to a fast-moving school AI landscape

The timing is not accidental. The announcement landed one week after New York City Mayor Zohran Mamdani and Schools Chancellor Kamar Samuels announced a one-year moratorium, effective in the 2026-27 school year, on student-facing generative AI for children in 2-K through 8th grade, alongside new screen-time restrictions. The city said the policy would affect nearly 600,000 students, or about two-thirds of system enrollment. Earlier this year, New York City Public Schools had already released AI guidance built around academic integrity, student privacy, data security and the continued necessity of human judgment. (nyc.gov)

Other systems have also been tightening the rules around classroom technology. Los Angeles Unified has published new student screen-time guidelines and district AI guidance, while New York State United Teachers in May adopted a resolution calling for limits on screen time, no student-facing AI for pre-K through second grade, and educator-led, supervised use of AI in any grade. In their Sept. 9 release, AFT and Microsoft said similar framework language has already been adopted by Los Angeles Unified, NYSUT and New York City Public Schools. (edtech.lausd.org)

At the federal level, meanwhile, the picture is still unsettled. The White House’s March 2026 AI legislative recommendations urged Congress to affirm that existing child privacy protections apply to AI systems, including limits on data collection for model training and targeted advertising. That recommendation itself underscores the point school districts have been making: the policy conversation is moving, but many of the practical rules districts need are still being worked out. Contract language is becoming a workaround for that lag. (whitehouse.gov)

What this does — and does not — solve

The deal’s practical value is also its biggest limitation. It applies to Microsoft arrangements, not to the full sprawl of AI products schools use or are testing. Districts that rely on a mix of classroom apps, tutoring tools, assessment systems, productivity platforms and browser-based AI assistants will still have to negotiate separately with other vendors or build these terms into their own requests for proposals and data privacy addenda. The memorandum explicitly says districts may request the standard’s protections for Microsoft educational AI products; it does not create a universal rule for the rest of the market. (aft.org)

It also leaves real work for district staff. Because the memorandum says protections can be incorporated into existing agreements and “need not be adopted in the identical form” of the standard, local legal and procurement teams will still need to read the amendments carefully. A district that assumes it is getting the full package could discover that the final addendum narrows some provisions, changes definitions or applies only to certain tools. That means school systems will need technical and legal capacity to compare what they asked for with what they actually received. (aft.org)

Still, the standard could matter beyond Microsoft if districts use it as a floor rather than a one-off deal. AFT’s fact sheet calls it a “gold standard,” and the memorandum says the principles are meant to be available to any district that wants them in its own purchases. If large districts begin asking other companies to match the same terms on training bans, tracking limits, deletion rights and family transparency, the center of gravity in school AI governance could move from guidance documents to procurement offices. That would be a consequential change: in education technology, contracts often shape day-to-day practice more than policy statements do. (aft.org)

The next thing to watch is not the announcement itself but adoption. Districts now have a template, but they will have to request it, negotiate it and enforce it. If enough of them do, Microsoft’s Sept. 9 deal with AFT and UFT may be remembered less as another AI principles rollout than as the moment school AI governance started showing up in contract redlines. (aft.org)