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 set of terms intended to limit how student and educator data can be used in AI systems. The agreement, announced Sept. 8, addresses a practical problem for districts buying classroom technology: privacy promises are common, but enforceable limits on vendors’ AI practices are far less common.
Under the framework described by the AFT’s announcement, student and educator data would not be used to train AI models, sold or repurposed. Schools would retain control over data use, retention and deletion, as well as when and how AI is deployed. The unions say districts can incorporate those protections into their Microsoft customer agreements, moving the issue from a broad policy statement into contractual terms.
What the proposed terms cover
The standard is broader than a ban on training a chatbot with student records. AFT says it includes requirements around safety safeguards, human oversight, security, accountability and responsible use. It also calls for accessible information to families about AI use, the data collected and the safeguards in place. Those provisions matter because school AI can range from an assistant that drafts classroom materials to tools that process pupil work, account information or other potentially sensitive data.
Details reported by the Associated Press, in a report published by ABC News, add two notable boundaries. The agreement includes third-party compliance audits, and it bars AI companions or features designed to create emotional attachment or dependency, or to keep a student engaged beyond a learning task. The latter provision reaches beyond conventional data privacy into product design: it targets systems whose interaction patterns could make a school tool behave more like a retention-oriented consumer service.
Neither source publishes the underlying contract language or specifies how an audit would be conducted, how frequently it would occur, or what remedy a district could seek after a violation. The announced standard is therefore a meaningful statement of intended terms, but not a public technical specification for Microsoft products. District procurement officials will still need to establish which services and accounts are covered, what data flows are involved, and how the relevant language appears in their agreements.
The AFT characterized the framework as legally enforceable once incorporated into a customer agreement. That is an important distinction from a government rule. It does not create a nationwide education-AI law, and its protections depend on the contractual relationship between a district and Microsoft rather than on a new federal regulatory regime.
Nationwide claim leaves a coverage question
Microsoft vice chair and president Brad Smith said the company would extend the agreement to school districts nationwide, according to the AFT. In the AP report, Smith said the standards would apply to all schools with Microsoft contracts beginning Nov. 1. Taken together, those statements point to an ambitious planned reach, but the public descriptions leave an implementation question unresolved.
The union’s primary announcement says districts can incorporate the protections directly into their Microsoft customer agreements. That wording suggests an adoption step by the district. The later AP account describes application to all Microsoft-contracted schools from Nov. 1. Neither public account establishes whether every existing school customer will receive the terms automatically, whether districts must amend or renew contracts, or whether particular products carry different conditions.
For schools, that difference is not merely administrative. A district that assumes coverage is automatic could miss an opportunity to verify its contract language; a district that believes a new negotiation is required may delay protections that Microsoft intends to make standard. The immediate test is likely to be the paperwork: whether the restrictions on training, sale, repurposing, retention and deletion appear in a form a district can enforce.
A vendor model, not an industry settlement
The agreement also illustrates the uneven bargaining position of individual school systems. Large districts may have procurement and legal staff capable of pressing technology vendors for detailed privacy terms. Smaller systems often do not. Elizabeth Laird of the Center for Democracy and Technology told the AP that a common framework could reduce the burden on districts negotiating with major companies one at a time.
Its broader reach remains unsettled. OpenAI told the AP it was working with the AFT and expected to finalize its own agreement, while Anthropic said it was working with the union on a safety and privacy standard. Those are discussions, not adopted terms. Google had not said whether it would provide comparable protections. That gap is consequential because Google Classroom is used by 150 million teachers and students worldwide, according to Google figures cited in the AP report.
The Microsoft agreement may give districts a concrete checklist for AI procurement: prohibit training on school data, preserve control over retention and deletion, require human oversight and family-facing disclosure, and scrutinize designs intended to deepen engagement. It does not establish that comparable protections follow students across the many services a school may use. As Nov. 1 approaches, the key evidence will be the contracts districts can actually obtain and enforce.
