Ten Rules and a 72-Hour Breach Clock: Microsoft's AI Standard for 50 Million US Students Goes Live Today

Microsoft's National AI Safety and Privacy Standard for K-12 education took effect on November 1, extending a set of contractually enforceable AI limits to every US school district the company works with — including districts that never asked for them. The standard, negotiated with the American Federation of Teachers and New York's United Federation of Teachers and announced on September 9, can be added to new or existing Microsoft agreements without renegotiation, and districts that write it into contracts gain the right to cancel and seek damages.

The document runs 30 pages and lists ten protections. Student, educator and customer data cannot be used to train or improve AI models, with a narrow exception for safety and security; it also cannot be sold, used for advertising or applied to product development. Students may not be tracked, AI companions and features designed to foster emotional attachment are banned, high-risk decisions require human review, breaches must be reported within 72 hours, and providers must keep required certifications, submit to third-party audits and fund an independent study on equity and accessibility. Families receive plain-language guides explaining how the AI products work.

"It is 30 pages, but in some ways it comes down to three words: privacy, safety, and transparency," Microsoft vice chair and president Brad Smith said at the launch. The company's fact sheet condenses the philosophy further: "AI supports, people decide." Two commitments survive the end of a contract — the bans on training and data sales are permanent, and deletion and breach-notification duties continue.

Microsoft is so far the only company to sign the unions' standard. AFT president Randi Weingarten called the agreement "a first of its kind" in September, arguing that federal and state laws lack enforceable guardrails for AI in schools, and said she hoped OpenAI and Anthropic would follow; both have said they are negotiating similar agreements. Google, which switched on its Gemini chatbot for many K-12 students through Google Classroom in August — a platform it says serves 150 million teachers and students worldwide — has not said whether it will offer comparable protections.

The standard lands in a fractured landscape. New York City and Los Angeles, the two largest US districts, have paused student use of AI for a year while they study how to incorporate it — New York's pause covers elementary and middle school, Los Angeles's extends through high school. Reaction outside the unions was mixed: Fairplay's Josh Golin warned a high-profile framework could be misunderstood as broader than it is, while the Center for Democracy and Technology's Elizabeth Laird said it makes progress on reducing inequity in AI protections between districts.

Alongside the pledge, Microsoft, OpenAI and Anthropic have contributed $23 million to the AFT's National Academy for AI Instruction, which aims to train more than 400,000 teachers to be AI-fluent by 2030. That dual role — funding teacher AI adoption while agreeing to hard contractual limits — is the trade the standard embodies: schools get enforceable terms instead of trust-me policies, and vendors get a template they hope will shape whatever rules come next.