Microsoft opens school AI privacy terms to U.S. districts

Microsoft made its school AI safety and privacy protections available for U.S. districts to request from November 1. Districts can add them to new or existing agreements without waiting for renewal or renegotiation, giving schools a contractual route to restrict how student and educator data is used.

The American Federation of Teachers, the United Federation of Teachers and Microsoft announced the standard on September 9. Microsoft reaffirmed the November 1 request date in an October 22 post. Each district must incorporate the protections into its agreement; availability does not automatically rewrite every classroom’s contract.

The commitments cover limits on AI training, bans on data sales and advertising, school control of retention and deletion, human oversight and information for families. AFT said incorporation makes the protections enforceable, with contractual remedies including termination for an uncured material breach and claims for damages.

The agreement covers authenticated use of generative AI products primarily designed and marketed for education. General-purpose productivity, search and cloud tools are excluded even when schools license them. Coverage follows the products and uses in each district’s agreement.

A narrow safety and security exception permits only necessary, purpose-limited processing, excluding general generative-model improvement.

Microsoft’s existing education AI application card already says customer content, inputs and conversations are excluded from foundational-model training. The contract standard adds enforceable rights and controls around data handling. The card also identifies grades, discipline, placement and admissions as consequential decisions outside the capabilities’ intended autonomous uses.

The agreement gives Speaker Coach and Speaker Progress a specific exception to prohibited practices under data minimization during user-initiated speaking activities, limited to the data needed for coaching and feedback.

Microsoft’s Learning Accelerators guide describes teachers configuring Teams speaking assignments, including practice time, attempts and feedback options, and tracking how students’ presentation skills develop over time. Speaker Progress transcribes speech and drafts feedback from that text. Educators review and can edit it before returning it. The application card warns that transcription quality, limited speech, dialects and disabilities can affect the feedback.

PowerPoint’s web-based Speaker Coach provides delivery feedback and optional body-language analysis. Its support page says body-language critiques are off by default and Microsoft saves no rehearsal video. These product explanations describe the inputs used in speaking practice and the review that accompanies the resulting feedback.

Review also varies by activity. The application card says Reading Coach stories reach learners without routine prior human review, making that experience unsuitable where every output needs advance expert approval.

The standard also covers encryption, access controls, independent testing and breach notification within 72 hours of awareness. Microsoft’s summary says training and sale bans are permanent, while deletion and breach duties continue after a contract ends. Families receive information about AI use, collected data and safeguards, and schools control material new features and data use, with data export and switching between providers also protected.