Microsoft Promises Not to Spy on Your Kid's Schoolwork

Microsoft just agreed to a set of ten privacy and safety promises for using AI in classrooms, signed with the American Federation of Teachers — the second-biggest teachers union in the US — and its New York City branch. The most important promise: Microsoft won't use students' or teachers' information to train its AI models. That matters because AI systems often learn by gobbling up data, and parents worry their kids' essays, grades, or personal details could end up inside a product used by millions.

The deal also limits how much data Microsoft collects in the first place, requires schools to explain to families in plain English how the tools work, bans "AI companions" (chatbots designed to act like a friend or confidant to a child), and requires a real human to sign off on any "high-risk" decision, like something affecting a student's record or placement.

Why now? New York City and Los Angeles schools both put one-year bans on student-facing AI, buying time to figure out proper guardrails. Microsoft seems to be reading the room: if more districts follow, it wants to be the company with rules already in place. Starting in November, school districts can bolt these terms onto new or existing contracts without renegotiating everything.

The union's president, Randi Weingarten, called the terms "iron-clad" and legally enforceable, saying nobody else — including the federal government — has done the real work. Tellingly, the same union has also accepted $23 million from Anthropic, Microsoft, and OpenAI to train teachers on AI. So this is less a rejection of AI in schools and more a fight over who writes the rules.