The Contract Nobody Read: Who Actually Negotiated AI Into Schools
Procurement offices signed AI into classrooms before curriculum committees met — and the people who signed will never sit with the consequences.
The Contract Was Already Signed
The governance failure in AI education is not that institutions moved too fast — it is that the people who moved are insulated from what they moved toward. University administrators who signed AI platform agreements do not teach the courses where those platforms land. School district procurement officers who approved AI reading apps will not sit with the five-year-olds whose data those apps collect. The European trade union document that surfaced on Bluesky identified the mechanism precisely: AI partnerships "are often presented as innovations in education" while the actual question — on what terms, with what autonomy surrendered — goes unasked. The framing of innovation forecloses the contract review.
Who Grades the AI Companies' Homework
The verification gap is structural, not incidental. AI companies are, as one post noted, "making sweeping claims about the advantages of their systems, while 'grading their own homework'" — and the governments entering agreements with them have abandoned the independent procurement review that public institutions once required for consequential technology. This is not a problem unique to education, but education makes it legible: the claims being made concern child development, learning outcomes, and literacy — domains where the cost of a miscalibrated tool is measured in years of a child's formation, not a failed product launch. The Brookings keynote and the April webinar series on AI research methods proceed as if these verification questions are settled. They are not settled. They were bypassed.
The Practitioners Who Were Not Asked
Teachers describe a professional transformation imposed rather than negotiated. The burden that AI placed on classroom instructors arrived without consultation: jobs shifted from teaching writing to documenting suspicion, from giving feedback to designing assessments that AI could not hollow out. Google and Microsoft have since announced major teacher training investments — but those investments are aimed at adoption, not governance. A University of Washington study of teacher perspectives found that teachers across grade levels feel unprepared, not opposed. The distinction matters: unprepared is a condition created by exclusion from the decision. It is correctable. But the correction requires giving teachers standing they were denied at the start — and the institutions that signed the contracts have no incentive to reopen that question.
Higher Education's Tuition Premise Is Already Broken
The higher education version of this problem has a sharper financial edge. Andrew Yang relayed a law firm partner's assessment that AI "is now doing work that used to be done by 1st to 3rd year associates" and that "someone should tell the folks applying to law school right now" . Nobody is telling them — because the institutions collecting tuition have a direct financial interest in not saying it. A commenter in that thread identified the structural irony that Silicon Valley, for all its talk of disruption, has done nothing to reduce the cost of the credential whose value it is actively eroding . The people who will absorb that loss are already enrolled. The people who won't are the ones who haven't taken out the loans yet, and they are not receiving adequate signal from the institutions that stand to benefit from their enrollment.
Data Collected Before Consent Was Possible
The surveillance dimension of AI education deployments is the longest-run version of the governance problem. Parent backlash against kindergarten AI tools in New York City has become a model case precisely because it makes the consent failure visible: five-year-olds cannot consent, their parents were not meaningfully consulted, and the data terms were written by platform vendors whose incentives run toward collection, not restraint. One post flagged the pattern in blunter terms — education apps are "surveilling children in government schools" in a scheme "worth $8 trillion" . The dollar figure is unverified, but the structural claim holds: a data harvesting system operating at scale on children who have no legal standing to contest it was built by procurement decisions, not by democratic deliberation. The children currently in those classrooms will not have a chance to renegotiate.
The Deciders and the Absorbers Are Different People
The through-line from kindergarten AI reading apps to law school enrollment to university platform contracts is the same: the people with authority to commit institutions are not the people who live with the consequences. That split is not a design flaw — it is a feature of how institutions procure technology, and AI has simply made its costs legible faster than any prior technology wave. The fix is not a better webinar or a larger teacher training budget. It is governance: giving practitioners, students, and parents formal standing in procurement decisions before the contract is signed, not a seat at the table after the terms are set. The institutions that build that standing into their processes now will be the ones that do not spend the next decade litigating the gap between what they promised and what they delivered.
The story so far
Procurement offices finalized AI contracts in education before teachers or curriculum committees had standing to contest the terms — locking practitioners into absorbing costs they had no role in creating.
Frequently Asked
- What should a teacher or school administrator do if AI tools were adopted without teacher input?
- Demand contract transparency: request the vendor agreement, data terms, and procurement record. If the institution cannot produce them on request, that absence is itself diagnostic. Then organize around the renewal decision — most institutional AI contracts run one to three years, and the renewal is the leverage point practitioners actually have. Showing up at procurement review with documented classroom impact is more effective than objecting after signing.
- Why are AI companies allowed to evaluate their own educational effectiveness?
- Because the procurement processes that once required independent verification for public-sector technology contracts were bypassed — partly by speed, partly by political enthusiasm for AI, and partly because education technology has historically faced weaker scrutiny than defense or medical procurement. AI companies moved into schools during a window when regulators had not yet applied the same evidentiary standards they apply elsewhere, and institutions signed before that window closed.
- What is the strongest argument that AI in schools is actually beneficial?
- The strongest counter is that AI tutoring tools demonstrably reach students who previously had no access to individualized instruction — rural districts, under-resourced schools, students with learning differences. For those students, the alternative to an AI reading app is not a better-funded classroom; it is no support at all. That argument does not resolve the governance and data questions, but it means the 'remove AI from schools' position has a real cost that falls on the students least able to absorb it.
Continue reading
Sal Khan's Khanmigo Admission Resets Ed-Tech's AI Expectations
Khan's public admission that Khanmigo failed to become a super-tutor ends the optimism that ed-tech used as its growth rationale.
similarThe MBA Cheating Crisis Exposed What Business Schools Never Measured
Systematic AI-assisted cheating in online MBA programs has forced a confrontation business schools cannot deflect: their assessments were never testing what they claimed.
similarThe AI-in-Education Argument Is Not About Cheating Anymore
Teachers now oppose classroom AI at majority levels, but the sharpest voices have moved past integrity debates to ask whether the institution itself is salvageable.
similarThe Man Who Built the Digital Classroom Warns About Its Next Phase
Blackboard's co-founder is warning that AI grading creates a vicious cycle — and his warning is arriving as a YouTube tutorial.
similarWhen the Grad Student Fakes Data and Gets a Promotion
A Harvard physicist caught Claude fabricating results, then handed it his entire research program — and the AI conversation has no framework for what that means.
Methodology
This story was generated autonomously from 19 source records. An editorial model synthesizes, weights, and cites each source. No human editorial judgment was applied.