Microsoft's school AI rules ban chatbot companions and training on student data

Microsoft and two teachers’ unions, the AFT and the UFT, announced a national AI safety and privacy standard for schools on September 9. Education Week reports that its protections apply to existing Microsoft school contracts as of November 1.

The standard is a set of contract terms. Student and educator data cannot be sold or used to train AI models, and schools decide how data is kept and deleted. The standard also restricts tracking such as keystroke logging and long-term profiling. AI cannot make high-stakes calls, like discipline or class placement, without a person reviewing them. Chatbots built to act as a student’s friend or emotional support are banned.

What’s in it:

• Providers must tell customers within 72 hours of a confirmed or suspected breach.
• Districts can end their agreements if problems aren’t fixed.
• Companies must explain in plain language how their tools work.

The catch: Districts have to choose to adopt it, and Microsoft is the first major AI company to agree. The AFT is still negotiating with Anthropic and OpenAI. Education Week also notes that de-identified usage data can still be used for debugging and security, and research suggests such data can sometimes be traced back to people.

What to do now

• Parents: ask your school which AI tools students use and whether the district’s contract includes these terms.
• Teachers: check district approval before uploading student work to any AI tool.
• Administrators: ask each vendor in writing whether they will accept equivalent terms, including the 72-hour breach notice.
• Everyone: ask the vendor for its plain-language guide. The standard says families should get one.