Microsoft announces AI in education policy with contract protection guidelines
On September 16, Microsoft released a policy that puts learner thinking and teacher judgment at the core of AI use in schools while adding safety and privacy clauses that districts can embed in contracts. The policy notes that AI‑generated answers may give the illusion of learning, so Copilot Chat for K‑12 is disabled by default and must be enabled by administrators based on student age.
Key points
- Microsoft policy disables Copilot Chat for K‑12 by default, requiring admin age‑based configuration.
- From November 1, U.S. districts can add data‑control, human‑verification, and feature‑oversight clauses to Microsoft contracts.
- Policy states learner and teacher data won’t train AI models, except for limited safety/security exceptions.
The accompanying agreement with the American Federation of Teachers, summarized for U.S. districts, allows schools to add protection clauses to new and existing Microsoft contracts starting November 1. These clauses cover school control over data retention and deletion, mandatory human verification for critical decisions, and oversight of any new feature or data‑usage changes. Microsoft also states that learner and teacher data will not be used to train its AI models, except for limited safety and security exceptions.
The article cautions that the provided documents are summaries, not full contracts, and that the terms do not automatically apply to schools outside the United States, such as in Japan. It also notes that the policy’s impact on learning effectiveness has not been independently verified.
What Should AI in Schools Protect: Learning Design and Contractual Terms
note.com · 22 September 2026
Loading the full article…
This text was published by note.com and written by 石川 明. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Policy & Regulation
All →- OpenAI says it plans to let third‑party groups conduct safety evaluations of its models · 1 src
- China probes DeepSeek, Moonshot AI over alleged Claude data routing · 3 src
- Anthropic CEO Dario Amodei urges global AI development pause over rogue agent risks · 34 src
- 66% of office workers use AI tools without company permission, PagerDuty survey shows · 2 src
- Donald Trump says US is officially renaming artificial intelligence to super intelligence · 1 src
Comments
via GitHub Discussions