Microsoft and teachers unions set enforceable AI safety and privacy standard for schools

Microsoft, the American Federation of Teachers and the United Federation of Teachers have announced a National AI Safety & Privacy Standard for schools, a contract-based framework meant to give districts enforceable protections when adopting AI tools. The September 9, 2026 announcement says U.S. school districts can incorporate the protections directly into Microsoft customer agreements, making the commitments legally enforceable rather than voluntary promises.

The agreement arrives while schools are still trying to decide how far AI should be allowed into classrooms. Some districts have moved toward restrictions, including New York City Public Schools' screen and AI limits for younger students and similar frameworks adopted in Los Angeles and by New York State United Teachers. Against that backdrop, the Microsoft agreement attempts to create a usable middle ground: AI tools can be used in education, but only with clearer rules on privacy, safety, transparency and control.

The strongest provision concerns data. The announcement says student and educator data cannot be used to train AI models, sold or repurposed, and that schools keep control over how information is used, retained and deleted. It also says students cannot be tracked. Those terms address one of the biggest worries among parents and teachers: that classroom AI systems could quietly turn learning records, questions, drafts or behavioral signals into training material or commercial profiles.

Safety rules are another focus. The standard says AI systems used in schools must be designed to avoid harmful or manipulative experiences, protect sensitive information, meet rigorous security expectations and preserve human oversight for consequential decisions. The agreement also responds to rising concern around companion-style chatbots by supporting a ban on social companion chatbots for students under 16, a position AFT President Randi Weingarten had previously outlined in a speech about devices and AI in early education.

Transparency is meant to make the rules visible to families and educators. Microsoft, AFT and UFT say companies should provide plain-language answers about how AI tools work, what data is collected and what safeguards are in place. Schools should retain control over when and how AI is used, while vendors communicate clearly about product changes, risks and controls. UFT President Michael Mulgrew said the standard gives districts tools to end agreements and seek damages if providers break the rules.

The agreement also highlights the vacuum left by the absence of comprehensive federal or state AI rules for schools. Existing privacy laws such as FERPA and HIPAA were not written for generative AI systems that can absorb context, respond conversationally and change quickly through product updates. The standard is therefore partly a procurement mechanism: districts can add enforceable language to technology contracts without waiting for a national law.

For the education technology market, Microsoft's move may set a template. Large vendors selling AI into schools will face pressure to make privacy and safety terms concrete enough for administrators, parents and unions to inspect. The standard will not resolve every classroom question about age, distraction, pedagogy or equity, but it makes one thing harder to avoid: AI systems used with children will increasingly be judged by enforceable governance, not just helpful demos.