A look at the playbook Amazon, Apple, Google, Microsoft, OpenAI, and others use to shape US schools to their benefit, and the pushback against unproven AI tools
One humid afternoon in the summer of 2025, representatives from Microsoft were waiting as some 200 teachers trooped into a conference hall in downtown Manhattan.
Context & Ripple Effects
The playbook now under scrutiny was built in public. In 2023, universities, companies, and nonprofits were already offering schools ready-made AI curricula while classrooms debated what to teach about AI's promise and peril (ready-made curriculums), and the industry-backed AI Education Project pushed literacy resources into low-income schools with backing from Microsoft, Google, and OpenAI. By late 2025 the effort had scaled: tech companies poured investment into K-12 and college adoption (AI adoption across US schools), pledged new commitments at a White House event alongside Amazon, IBM, Code.org, and others, and funded union-run training directly.
What has changed is the frame. The New York Times piece treats these moves not as philanthropy but as market-shaping — and documents the counter-current: educators questioning tools whose classroom value is unproven. That tension is already visible internationally, where governments are racing to deploy GenAI in schools even as UNICEF urges caution.
First-order effects
- Microsoft, OpenAI, and Anthropic's funding of American Federation of Teachers hubs targeting 400,000 teachers ($M to the AFT for AI training hubs) puts vendor-aligned curriculum directly in front of the people who decide what enters classrooms — while those same teachers are now the audience for coverage challenging whether the tools work.
Second-order effects
- Unions have become a distribution channel, which forces rival labs to compete through the same gatekeeper: more money flows to AFT-style partnerships rather than direct district sales, shifting edtech competition toward whoever trains the trainers.
Third-order effects
- If 'unproven' becomes a procurement criterion, expect calls for independent evaluation standards for classroom AI — mirroring the international split where deployment races ahead of bodies like UNICEF urging restraint, and embedding whichever vendors survived early adoption deep into school infrastructure.
The trend: AI labs are moving from selling software to schools to funding the training pipeline that decides which tools schools trust — and scrutiny of unproven products is becoming part of that pipeline.