Microsoft has new AI privacy rules for schools

Microsoft agreed with the American Federation of Teachers and New York City's United Federation of Teachers on ten principles for AI in schools that districts can write into contracts. Commitments include not training on student or educator data, limiting collection, plain-language disclosures for families, prohibiting AI companions, and requiring human review for high-risk decisions.

The deal arrives one week after New York City and Los Angeles announced one-year bans on many student-facing AI tools while they figure out guardrails. Starting in November, districts can add the Microsoft terms to new or existing agreements without renegotiating entire contracts.

AFT President Randi Weingarten called the provisions iron-clad and said no one else—including the federal government—has done the real work. The union has also backed educator AI training, including a $23 million hub funded by Anthropic, Microsoft, and OpenAI.

We're reading this as market adaptation under political pressure, not a sudden conversion on youth privacy. When two of the country's largest school systems hit pause, vendors that want classroom share either accept enforceable limits or watch the ban wave spread. Microsoft chose the contract path.

That's the useful part. Soft pledges and marketing PDFs don't survive a superintendent's legal review. Terms districts can attach without a full renegotiation lower the activation energy for privacy baseline. No training on student data, no AI companions, human review for high-risk decisions—these are the fights parents and teachers were already having in public meetings.

Eyes open on the gaps. Opt-in by district means uneven protection; kids in systems that never adopt still sit on whatever default Microsoft shipped. "High-risk" will be argued case by case. And the same union criticizing student-facing AI before middle school is co-building training infrastructure with the labs—so this is coexistence politics, not a full stop.

I'm glad useful classroom AI has a shot at surviving if it can clear a privacy floor. The trajectory if this becomes normal is vendor competition on enforceable education terms, not just feature demos. If Microsoft's rivals don't match, expect more district-level bans instead of negotiated guardrails.

I'm watching November uptake numbers, whether Google and OpenAI publish comparable bindable terms, and whether the NYC/LA pauses end with adoption under these rules or harden into longer bans.

Parents and some educators have pushed back on AI and screens in schools more broadly. Weingarten has called for bans on screens before third grade and student-facing AI before middle school, while still engaging companies on teacher preparation for the tools that remain.