Three Rivals, One Classroom: AFT's AI Coalition

Watch how three companies behave and you learn more than any of them will tell you. In almost every market that matters — enterprise seats, developer mindshare, model benchmarks — OpenAI, Microsoft, and Anthropic are locked in the kind of competition that ends careers and reshapes org charts. So it's worth pausing on the one room where all three showed up together, checkbook in hand, to fund the exact same thing: a teachers' union program to train 400,000 K-12 educators on AI. When fierce rivals agree to co-sponsor a shared referee, the interesting question isn't why they cooperated. It's what each of them is quietly buying.

The deal, on paper

In July 2025, the American Federation of Teachers announced the National Academy for AI Instruction, a five-year initiative to bring practical AI training to educators across the country. The AFT — the second-largest teachers' union in the U.S., with roughly 1.8 million members — is the anchor. Its New York City local, the United Federation of Teachers, provides the physical home: a flagship training hub inside the UFT's Manhattan headquarters, which opened in Fall 2025, with more hubs planned across the country by 2030.

The money is the tell. The academy is backed by $23 million in total funding, and the donor list reads like a truce. As Education Week reported, Microsoft, OpenAI, and Anthropic are all founding partners. OpenAI's slice alone is $10 million over five years — $8 million in direct funding and $2 million in in-kind technical resources. The stated goal is concrete and large: train 400,000 educators over five years, roughly one in ten of the nation's K-12 teachers.

Three companies that will not agree on a shared benchmark, a shared safety standard, or a shared definition of "AGI" agreed, without apparent friction, to fund the same union's classroom. That's not generosity. That's positioning.

The framing from the AFT side is member-first: get ahead of a technology that is already in classrooms, on the union's terms, with training the union helps design rather than tools dropped on teachers from above. TIME's coverage captured the pitch — teachers shaping how AI enters their profession instead of being shaped by it. That's a real and defensible motive. But it doesn't explain why three competitors would all want a seat at that particular table.

Why competitors fund the same referee

Start with the obvious: none of these companies is doing this for the $23 million to come back as revenue. The academy is not a sales channel. What it is, for each vendor, is a position in the room where the next generation of professional norms gets written.

K-12 education is a market with a peculiar shape. It's enormous, it's slow, and it's sticky. A teacher who learns to plan units, draft parent emails, and build quizzes with one company's assistant doesn't casually switch next semester — the workflow becomes muscle memory, and muscle memory is the most durable moat in software. The company that trains the teacher isn't selling a subscription today. It's setting a default for a decade. That's the same logic behind OpenAI making its flagship model free to verified teachers through 2027: the giveaway isn't charity, it's habit-formation at the front of a very long funnel.

So why cooperate instead of racing? Because the alternative to a shared, union-blessed academy is worse for all three. Absent a neutral venue, AI in schools becomes a series of district-by-district knife fights — procurement battles, pilot programs, competing PD vendors, and a growing chorus of educators who distrust the whole category. A jointly funded academy converts a messy, adversarial rollout into a legitimized one. It hands each vendor the credibility of union endorsement, dilutes the appearance of any single company "capturing" the classroom, and — crucially — establishes that AI-in-education is inevitable and normal, which is the precondition every one of them needs before the market can grow at all. They're not competing to own the room yet. First they're all funding the construction of the room.

What the union gets, and what it risks

The AFT's calculation is easier to read, and more sympathetic. Its members are already fielding students who cheat with ChatGPT, already being handed AI tools by administrators, already anxious that the technology is a threat to their craft and their headcount. A union that ignores this looks asleep. A union that only resists looks Luddite. The academy is the third option: engage, train, and try to steer.

But the coalition carries a real conflict of interest, and it deserves to be named plainly rather than waved away. When the companies selling AI tools are also funding the union program that teaches educators to trust and adopt AI tools, the training is not a neutral evaluation — it's downstream of the sponsors. Word In Black raised exactly this concern, noting that vendor money flowing into union training invites questions about whose interests the curriculum ultimately serves. And the dissent isn't only external: a rank-and-file caucus within the Oregon NEA has publicly opposed what it calls the "AI grift," arguing that unions should be protecting members from an overhyped technology, not evangelizing it on the vendors' behalf.

Both things can be true at once. The academy can genuinely help teachers who would otherwise be left to figure this out alone, and it can function as a reputation-laundering and default-setting exercise for three companies with billions riding on the education market. Adults should be able to hold both. The training a teacher receives at the NYC hub may be excellent and the funding structure may still tilt the frame — those are separate audits, and this series runs both.

The pricing context nobody mentions in the press release

Here's the part the announcements leave out. In the enterprise market these same companies fight over, AI assistants are not free. As of 2026, Anthropic's Claude for Team runs about $20 per user per month, Google's Gemini for Education bundles start around $14, and Microsoft's Copilot for Education sits near $21 per seat with deep Office 365 integration. Against that backdrop, the academy — free training — and OpenAI's free-for-teachers offer look less like philanthropy and more like customer acquisition in a segment where every vendor knows the paid conversation is coming later.

That doesn't make the training worthless. It makes it legible. A teacher walking into the academy should understand that the room was built by companies that will, eventually, want the district's purchase order — and should judge the curriculum accordingly. The best defense a teacher has against being positioned is knowing that positioning is happening. The three rivals shared a classroom for reasons that have very little to do with the classroom. Knowing that is the first lesson worth teaching.

Part 61 of 100 in the ChatGPT for Teachers series. Previously: The NYC Hub Training America's Teachers on AI. Next: What OpenAI's Free Teacher AI Course Covers. Browse more builder insights or explore AI skills for education at aiskill.market.