Microsoft, AFT Agree Student AI Privacy Rules

Microsoft and the American Federation of Teachers reached a binding agreement to bar student and educator data from AI training, require audits and plain-language disclosures; other vendors have not joined.

Microsoft and the American Federation of Teachers reached a legally binding agreement last week that limits how the company can use data from students and educators in schools that contract with Microsoft.

The contract bars using student or educator data to train AI models, with a narrow exception for information needed to protect student safety. Microsoft committed not to sell student data, use it for advertising, or apply it to product development.

The agreement requires third-party audits to verify compliance and plain-language disclosures to families explaining how tools work. It also forbids AI companions or features designed to foster emotional attachment or to keep students engaged beyond a learning task.

Microsoft’s vice chair and president, Brad Smith, described the standard as a high bar for child privacy and AI safety and noted the company will apply the terms to every district with a Microsoft contract. The provisions will take effect for districts with Microsoft contracts beginning Nov. 1.

AFT President Randi Weingarten described the pact as the first of its kind and framed it as returning control of students’ education to parents and educators rather than ed-tech firms.

The agreement follows one-year moratoriums announced by New York City and Los Angeles on student use of AI while those districts audit their education technology contracts and review how to integrate the tools. New York City’s pause covers elementary and middle schools; Los Angeles extended its restriction through high school.

OpenAI and Anthropic have expressed support and said they are discussing similar pacts with the union. Google has not publicly committed to comparable limits. In August, Google expanded access to its Gemini chatbot for many K-12 users of Google Classroom, a change that prompted privacy criticism.

Elizabeth Laird of the Center for Democracy and Technology noted the agreement shifts the burden of privacy protections onto technology companies and could help districts that lack bargaining power secure stronger terms. She also pointed to the pact’s limits on use of sensitive data such as gender, race, age and ZIP code.

Josh Golin, executive director of Fairplay, warned the pact does not resolve whether AI should be used in classrooms or at what age, and cautioned that robust privacy rules might be read by some districts as an endorsement to deploy AI widely.

The AFT launched a National Academy for AI Instruction last year with $23 million in funding from Microsoft, OpenAI and Anthropic to train teachers on classroom uses of AI.

Districts are beginning audits of education technology and companies are weighing whether to adopt similar agreements.

Articles by this author