🔍 Read the full analysis: The New Rules For Human Oversight In School AI, Explained on ThorstenMeyerAI.com
TL;DR
Microsoft says U.S. school districts can use a contract from Nov. 1, 2026, that bars covered education AI products from training on student data, tracking students, acting as companions or making decisions without meaningful human review. The agreement is contractual, not a new law, and its scope and how districts adopt it are not fully clear. Separate California and New York laws set limits in workplaces, health care and chatbot products for children.
Microsoft says U.S. school districts can use a new 31-page contract from Nov. 1, 2026, that limits how covered education AI products handle student information and requires meaningful human review before decisions about students or teachers. The agreement is a procurement tool, not a federal law, and applies to education products within its scope rather than every AI service used by schools.
The document, called the National AI Safety & Privacy Standard, was announced Sept. 9 by the American Federation of Teachers (AFT), its United Federation of Teachers affiliate and Microsoft. It is a memorandum of agreement between Microsoft and the National Academy for AI Instruction, an educator-led training body announced by the AFT in 2025. The document also has blank signature slots for other providers, according to the source report.
Its terms prohibit covered providers from using prompts, answers, uploaded files, behavioral signals or metadata from students and educators to train, fine-tune or otherwise improve AI models. The document says de-identification is not an exception to that ban. It allows limited processing to detect specified safety threats, including self-harm risk, grooming, threats of violence and security attacks; that data cannot then be used for general model improvement.
The agreement also restricts location and behavioral tracking, keystroke logging, continuous passive attention monitoring, biometrics and emotional inference. It bars companion-style features designed to foster emotional attachment or dependency. Covered systems must not make decisions about students or teachers without meaningful human review or prior approval. Some monitoring prohibitions include an exception for explicit written approval from the school customer, and agent-like features that act outside a provider’s system are to be off for students unless an administrator enables them.
Infographic 01 · The lines drawn this autumn
Seven weeks, one direction: a human stays in charge
From the school contract to state laws: what was signed when, and what the classroom standard actually requires.
- 2 Feb 2025EU: emotion inference banned at work and schoolAI Act Art. 5(1)(f), except medical or safety reasons
- 9 Sep 2026AFT, UFT and Microsoft announce the school standardA contract districts can add to their Microsoft agreements
- 30 Sep 2026California signs AB 1883, SB 947, SB 951, AB 1979Emotion AI at work, robot bosses, AI layoff notices, AI in clinics
- 28 Oct 2026New York signs its chatbot law for minorsS.9051B/A.10379C: no unsafe companion features for under-18s
- 1 Nov 2026Microsoft makes the standard available to all US districtsPer Microsoft’s fact sheet; districts “can add them”
- 1 Jan 2027Most California rules take effectAB 1883, AB 1979, SB 951
- 1 Jul 2027SB 947 becomes operativeHuman corroboration before AI-driven discipline or firing
What Microsoft’s classroom AI may not do
- Learn from students or teachers. No training on their data; de-identification is “not a loophole”.
- Track them. No keystroke logging, attention monitoring, long-term profiling or biometrics.
- Befriend them. No companion features designed to foster emotional attachment.
- Decide alone. No decisions without meaningful human review or prior approval.
Where it is narrower
- Scope: education products only. General-purpose productivity and workplace tools are excluded by definition.
- Opt-in or automatic: Microsoft says districts “can add” the terms; the agreement says “upon request within 90 days”.
- Exceptions: a narrow safety exception for training; the no-tracking list allows written approval by the school.
“Anything less than legally enforceable provisions is simply a wish list.”Randi Weingarten, AFT President, 9 September 2026
Sources: AFT/UFT/Microsoft release, 9 Sep 2026 · Standard MOA v1.0 · Microsoft fact sheet · Governor of California, 30 Sep 2026 · Governor of New York, 28 Oct 2026 · EU AI Act, Art. 5(1)(f). Effective dates for California follow the chaptered texts and the state’s default start date for regular-session bills.
Powered by Thorsten Meyer AI
How District Contracts Could Shape AI Use
The immediate effect depends on whether districts adopt the terms in their own Microsoft contracts. The agreement says districts can incorporate its provisions into those contracts; the source report says that would give districts contractual remedies, including ending a contract or seeking damages for a breach. That makes procurement decisions central: the standard does not itself create a general legal duty for every school or AI company.
For students and families, the proposed boundaries address data use, monitoring and emotionally oriented chatbot features, as well as who makes consequential decisions. For educators and administrators, the human-review requirement is intended to keep software from making certain decisions on its own. The agreement also directs the Academy to train educators not to use AI for automated grading without review, discipline, placement or emotional assessment.
These terms sit alongside state measures rather than replacing them. California has enacted restrictions concerning workplace emotion and neural-data monitoring, algorithmic employment decisions and AI’s role in health care. New York has enacted rules targeting unsafe chatbot companion features for children and unverified users. Together, the developments put human oversight and limits on sensitive data into contracts and laws, but their coverage and enforcement routes differ.
Popular right now
Philips 2200 Series Fully Automatic Espresso Machine with Classic Milk Frother
Vitamix Propel Series 510 Professional-Grade Blender, 3 Programs, Black
Ninja Foodi 8-Quart DualZone 6-in-1 Air Fryer, Grey (DZ201)
As an affiliate, we earn on qualifying purchases.
Three Different Routes to Oversight
The school agreement was announced by the AFT and Microsoft on Sept. 9, 2026; Microsoft says it became available to school districts on Nov. 1. It is framed as a contract that a district may put into its Microsoft agreement, not a universal rule imposed by government. The source describes it as covering products designed and marketed for education and excluding general-purpose productivity and workplace tools.
California Gov. Gavin Newsom signed four measures on Sept. 30, according to the supplied report: AB 1883 restricts AI emotion-reading and neural-data collection at work; SB 947 limits discipline or firing decisions made by algorithms alone; AB 1979 restricts AI from performing health-care functions reserved for licensed professionals; and SB 951 requires layoff notices to disclose when AI replaced jobs. Most take effect Jan. 1, 2027; SB 947 is scheduled to take effect July 1, 2027.
New York Gov. Kathy Hochul signed a law on Oct. 28 addressing chatbot features for minors and unverified users. The supplied account says it prohibits unsafe companion features, including systems that pretend to be human or encourage emotional dependency. Unlike the school memorandum, the California and New York measures are laws. Their provisions and dates should not be treated as interchangeable with the contract’s terms.
„anything less than legally enforceable provisions is simply a wish list.“
— AFT President Randi Weingarten
Adoption and Coverage Remain Open
It is not clear from the supplied material whether the agreement applies automatically to every eligible Microsoft education product or only when a district incorporates its terms into a contract. The source itself says Microsoft’s description, the agreement and press reports differ on that point. District-by-district adoption will determine how widely the contractual protections operate.
The agreement is described as covering AI products designed and marketed for education, while excluding general-purpose productivity and workplace tools. How that distinction applies to products used in mixed settings, or to features that change over time, is not specified in the supplied account. The source also notes written school approval can create exceptions to some monitoring restrictions.
Details of enforcement in practice remain unreported: the source does not identify districts that have adopted the contract, describe any breach dispute, or give information about how often the safety exception might be used. The state laws’ implementation and enforcement details are also not set out here. The claims about what each measure requires should be checked against the full legal texts and contract before drawing conclusions about a particular product or school.
District Procurement Is the Next Test
School districts can review the memorandum and decide whether to write its terms into their Microsoft contracts. Their procurement teams will need to determine which products and features are covered, how human review is documented, and whether any permitted exceptions have been approved in writing. The agreement specifies breach reporting within 72 hours and deletion of data from active systems within 180 days; those are contract terms, not evidence that a violation has occurred.
As California’s provisions approach their 2027 start dates, employers and health-care organizations will need to examine which tools and decisions fall within each law. New York’s chatbot restrictions will likewise require attention from providers and organizations offering relevant services to children. The next concrete evidence of impact will be district contract adoption and implementation, alongside official guidance or enforcement information from state authorities.
Key Questions
Is the Microsoft school AI standard a law?
No. It is a memorandum of agreement that districts can incorporate into Microsoft contracts. Its reach depends on the contract and products covered; it is not described as a law applying automatically to every school or AI provider.
What does the agreement say about student data?
For covered education products, it bars using student and educator prompts, files, behavioral signals and metadata to train or improve AI models. It allows limited processing for specified safety threats, but says that information cannot be used for general model improvement.
Does it cover every AI tool used in a school?
No, not according to the supplied account. It covers products designed and marketed for education and expressly excludes general-purpose productivity and workplace tools. How mixed-use products are treated is not clear from the available description.
What does human review mean under the agreement?
The memorandum says covered systems may not make decisions about students or teachers without meaningful human review or prior approval. The supplied material does not spell out one universal review process; districts and providers would need to apply the requirement in their contracts and practices.
When do the related state laws take effect?
According to the supplied report, most of the listed California provisions take effect Jan. 1, 2027, while SB 947 takes effect July 1, 2027. New York’s chatbot law was signed Oct. 28, 2026; the source material does not give its effective date.
Source: ThorstenMeyerAI.com
