🔍 Read the full analysis: Rules That Keep Humans In Charge: What Microsoft’s School AI Pact And California’s New Laws Actually Say on ThorstenMeyerAI.com
Get the latest gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
Microsoft says it will make a school AI contract standard available to U.S. districts starting November 1, 2026. The agreement restricts uses of student data and tracking and requires human review for covered decisions, but it is a contract framework, not a law, and applies to education-designed products. California and New York have adopted separate rules affecting workplace, health care and child-facing AI, with implementation details and effective dates varying by law.
Microsoft says that starting November 1, 2026, every U.S. school district can use a contract standard limiting how covered education AI products handle student data, track children and make decisions. The 31-page agreement is not a federal law or a blanket Microsoft policy: districts must put its terms into contracts for the protections to be enforceable, and the standard covers products designed and marketed for education.
The agreement was announced on September 9 by Microsoft, the American Federation of Teachers and the United Federation of Teachers. Its contracting parties are Microsoft Corporation and the National Academy for AI Instruction, an educator-led training body announced by the AFT in 2025. The document also has empty signature slots for other AI providers. Microsoft President Brad Smith described it as setting a high bar for child privacy and AI safety.
Under the text, providers may not use prompts, answers, uploaded files, behavioral signals or metadata from students, teachers and administrators to train, fine-tune, benchmark or otherwise improve AI models. The agreement says de-identifying the data does not exempt it from the restriction. A narrow exception permits processing for specified safety concerns, including self-harm risk, abuse material, grooming, threats, bullying and security attacks; that data may not then be used for general model improvement.
The terms also bar specified forms of tracking and emotional inference, including continuous passive attention monitoring, long-term profiling and biometric data collection. They prohibit companion-style features designed to encourage emotional attachment or dependency. Covered systems cannot decide matters concerning students or teachers without meaningful human review or prior approval. Some tracking restrictions allow exceptions with a school’s explicit written approval; the source material does not reproduce the full agreement’s clauses or define every approval process.
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 Enforce Limits
The agreement’s practical force comes from school procurement, not a new government regulator. A district that incorporates the terms into its Microsoft contract can use that contract to hold the provider to the commitments; the source says districts could end a contract and seek damages for breaches. That makes the scope and wording adopted by each district consequential. Merely making a template available does not establish that every district has signed it or that every Microsoft product is covered.
For educators and families, the listed boundaries address several points of direct concern: whether student work feeds model development, whether a system monitors behavior or infers emotion, and whether software can make consequential choices without a person reviewing them. These protections may also influence purchasing discussions with other vendors, though the agreement itself does not bind providers that have not signed or districts that do not adopt its terms.
The development sits alongside state laws addressing related risks in other settings. California’s new statutes target particular workplace and health-care uses, while New York’s law addresses certain chatbot features for minors and unverified users. Together, the measures show different routes for setting limits—private contracts and state legislation—but they do not create one uniform national AI rule.
Products Worth Considering
Logitech MX Brio Ultra HD 4K Webcam, Graphite
Samsung Galaxy Tab S10 FE 128GB Wi-Fi Android Tablet, Blue
Samsung 27 Odyssey OLED G5 (G50SF) QHD QD-OLED Gaming Monitor
As an affiliate, we earn on qualifying purchases.
Three Rules, Different Legal Reach
California Governor Gavin Newsom signed four AI-related bills on September 30, 2026, according to the source. AB 1883 restricts AI systems that read workers’ emotions or collect neural data. SB 947 bars an algorithm from making a firing or disciplinary decision on its own. AB 1979 limits AI from performing health-care work that law reserves for licensed professionals. SB 951 requires layoff notices to say when AI has replaced jobs. The source says most take effect January 1, 2027, with SB 947 taking effect July 1, 2027.
New York Governor Kathy Hochul signed a law on October 28, 2026, addressing chatbots that interact with children. The source describes it as prohibiting unsafe companion features, including systems that pretend to be human or foster emotional dependency, for minors and unverified users. The provided material does not give the statute number, full statutory text or effective date.
The Microsoft standard has a different reach: it is a contract framework for education products, rather than a law applying across an industry. The source says general-purpose productivity and workplace tools are expressly excluded. It also reports that educator guidance under the agreement discourages automated grading without review and AI use for discipline, placement or emotional assessment. Those points should not be confused with a universal legal ban on such uses.
“Anything less than legally enforceable provisions is simply a wish list.”
— AFT President Randi Weingarten
Adoption and Coverage Still Vary
It remains unclear from the source whether Microsoft makes the terms apply automatically to existing customers or only when a district requests and incorporates them. The material itself describes the rollout differently in places, while also saying districts can write the provisions into their contracts. The agreement’s availability should not be treated as proof that every district is covered.
The standard’s detailed operation will depend on the contract language, product scope and any written exceptions approved by a school. The source does not identify participating districts, report signed contracts, or provide evidence of enforcement or compliance. It also does not establish how the safety exception will be audited beyond a requirement for a senior officer to confirm specified handling in writing once a year if asked.
For the state measures, the source lists California’s effective dates but does not supply full statutory text or explain enforcement procedures. It gives fewer details for New York, including no effective date. The laws’ precise reach and how agencies or courts apply them remain matters for the actual statutes and subsequent implementation.
Districts Prepare for Contract Reviews
Starting November 1, districts can seek to use the Microsoft standard in their contracts. The next practical step is whether local school systems adopt the language, which products they treat as education AI, and how they handle any allowed exceptions. Districts and providers may also face questions about breach reporting, deletion timelines and data location: the source says breaches must be reported within 72 hours, data deleted from active systems within 180 days, and data kept in the EU or North America.
California’s rules are scheduled to begin taking effect in stages from January 1, 2027, followed by SB 947 on July 1. Employers and health-care organizations will need to determine which tools and decisions fall under each law. New York’s implementation timeline cannot be stated from the supplied material; the next step is clarification through the law’s text and state guidance.
For organizations reviewing AI systems, the source points to practical areas to check: features labeled sentiment, mood or engagement; processes where software influences discipline or termination; and whether a human has meaningful authority to review consequential decisions. Those are reported recommendations, not a substitute for legal advice or proof that a particular tool violates a rule.
Key Questions
Does Microsoft’s agreement automatically cover every U.S. school district?
No. Microsoft says districts can use the standard, but the source says its protections are made enforceable by incorporating the terms into district contracts. It does not confirm that every district has adopted them.
Which AI products does the school agreement cover?
It covers products designed and marketed for education. The source says general-purpose productivity and workplace tools are expressly excluded.
Does the agreement ban all processing of student data?
No. It bars using student and educator data to train or improve AI models, while allowing a narrow exception for specified safety and security concerns. The source says data handled under that exception cannot then be used for general model improvement.
When do California’s laws take effect?
The source says most of the four laws signed September 30, 2026, take effect January 1, 2027. SB 947 is scheduled to take effect July 1, 2027.
What does the source say about New York’s chatbot law?
It says Governor Kathy Hochul signed the law on October 28, 2026, and describes restrictions on unsafe companion features for minors and unverified users. The supplied material does not give the law number or effective date.
Source: ThorstenMeyerAI.com
