This week, The New York Times published Natasha Singer's "How Big Tech Captured American Schools," an account of how a handful of technology giants came to influence "nearly every step of the education supply chain" — the laptops, the apps, the teacher trainings, the curricula, even the state laws that decide what children must learn. One retired educator quoted in the piece put it in a single line:
"We aren't sitting at the table. We're the meal."
Lois Weiner, education professor emerita — The New York TimesShe was talking about a $23 million deal in which AI companies fund the training of the very teachers who will decide whether to adopt their products. But she could have been describing the entire architecture of educational technology as it exists today — because the problem isn't any one company. The problem is the business model.
Free is a price, and schools have been paying it for a decade.
The article traces how "free" apps and low-cost laptops won the classroom: districts got tools, and in exchange, companies got something more durable — default status, data, and what one framing in the piece calls customers for life. Ten years and tens of billions of dollars later, the Times reports, districts are experiencing what Singer calls "Chromebook remorse," and the same companies that sold schools on coding are now selling them on AI — with curricula, trainings, and explainer videos featuring cartoon ducks.
Educators are right to be skeptical this time. But skepticism alone isn't a strategy. Schools still need AI — for lesson differentiation, for administrative load, for accessibility, for the genuine work of teaching. The question was never whether AI belongs in education. It's who owns it when it arrives.
There is another way to do this. It's called sovereignty.
At MyAuX Labs, we build bespoke AI models that institutions own outright and run on their own hardware. Not a subscription. Not an API key. Not a "free tier" that matures into a dependency. A model — built for your institution, running inside your walls, answerable to no one but you.
For schools and educators, that architecture changes everything the Times article warns about.
Student data never leaves the building.
A sovereign model runs on hardware the district owns. Student essays, IEP notes, behavioral records, parent emails — none of it transits a third party's cloud, trains a third party's next model, or feeds a third party's advertising engine. There is no downstream. The Times reminds us that one major classroom provider paid $170 million over the harvesting of children's data. With a sovereign model, that category of risk doesn't get "mitigated." It gets deleted.
No hidden curriculum.
When a trillion-dollar company writes the coding lessons, funds the teacher trainings, and supplies the AI tutor, the worldview comes bundled. As the Times put it: there is no Raytheon AP Physics. A sovereign model carries no brand to launder and no product funnel to fill. It teaches what your educators tell it to teach — your standards, your pedagogy, your community's values. The district writes the curriculum. The model serves it.
Teachers become owners, not end-users.
The article's most uncomfortable comparison is to pharmaceutical reps training doctors. That dynamic exists because teachers are treated as an adoption channel — the last mile of someone else's sales motion. Sovereignty inverts it. When the district owns the model, teachers aren't being trained on a product; they're shaping their tool. Feedback doesn't disappear into a vendor's roadmap. It becomes the next version.
Traceability instead of trust-us.
Our context-constrained architecture is built for traceability — the ability to see why the model produced what it produced. In a classroom, that isn't a nice-to-have. An educator who can't interrogate a tool can't responsibly put it in front of children. "Technology civics" — the discipline the article describes, teaching students to ask who built this, who profits, how does it affect me? — should apply to the school's own AI first. A sovereign, inspectable model is one your students can actually study, not just consume.
No customer-for-life pipeline.
The economics of Big Tech in schools depend on the relationship outliving graduation — hook students on an ecosystem at eight, monetize them at twenty-eight. A model your district owns has no such ambition. It has no growth targets, no engagement metrics, no reason to make itself sticky. It's infrastructure, like the building. It works for the school because the school is not the product — it's the proprietor.
Budget certainty in an era of rented everything.
Districts have watched "free" become freemium become line-item. Owning your model means the cost is knowable, the capability is permanent, and no pricing change or terms-of-service update three time zones away can hold your instructional tools hostage in October.
The Times article ends with a middle-school teacher in Oakland teaching her students to decode the technology in their lives instead of being coded by it. That's exactly the right instinct — and it deserves infrastructure to match. A school that rents its intelligence from the companies it's teaching students to scrutinize is arguing against itself.
Own your model. Own your data. Own the future you're preparing students for.