Teachers Are Getting AI Training. Their Union Took $23 Million From Big Tech To Pay For It.
The American Federation of Teachers wants to help its members understand and manage AI in classrooms. Where the money came from is harder to defend.
Key points
- The American Federation of Teachers, one of the largest US teacher unions, accepted $23 million from AI companies to fund classroom AI training programmes.
- New York City teachers attended hands-on sessions in 2025 designed to teach both how AI works and how to stop students from using it to avoid thinking.
- The funding arrangement has drawn criticism from educators and labour advocates who question whether a union should take money from the industry it is meant to hold accountable.
- Some teachers view AI as a direct threat to learning and want training that helps them resist it, not accommodate it.
Darius Saczuk does not trust AI. The New York City high school teacher spent a day earlier this year in a windowless downtown Manhattan conference room, surrounded by dozens of colleagues, learning to use it anyway. His logic is blunt: know your enemy.
The training was part of a push by the American Federation of Teachers (AFT), a union representing roughly 1.7 million members across schools and public services, to prepare its members for a classroom where AI tools like ChatGPT are already present whether teachers want them or not.
Who is paying for this?
The $23 million comes from AI companies, and that fact has caused friction. A labour union's job is to protect workers from harm, including the kind new technology can cause. Taking large cheques from the firms building that technology sits uneasily with that role.
The AFT says the money lets it run training at a scale it couldn't otherwise afford, and that teachers will encounter AI regardless. Those sceptical of the arrangement argue that industry cash can quietly shape what the training covers and what it leaves out. Dave Eggers made a version of this point directly to OpenAI staff in July, telling around 200 employees what he believed they had done to teachers and students.
What does the training actually cover?
The sessions give teachers two things: a working understanding of how AI tools function, and practical strategies for keeping students from using those tools as a shortcut instead of actually learning.
For Saczuk, the second part is what counts. His worry isn't that AI is complicated. It's that a large language model, the kind of AI that can write an essay or solve a maths problem on request, does exactly what a struggling student might want: produces a finished answer with almost no effort from the person asking.
What does this mean for students and parents?
The training targets teachers, not classrooms directly. But the downstream effect lands on students. A teacher who understands how AI generates text is better placed to spot a submitted essay that a student didn't write, and to design assignments that are harder to hand off to a machine.
Our earlier story on AI exam proctoring going wrong at Mexico's UNAM, where 58,000 students faced a retake after automated invigilation produced anomalous results, shows how quickly AI tools in education can go sideways without human oversight. The AFT's bet is that informed teachers are the most reliable check.
Parents should expect schools to keep wrestling with this for years. There's no settled answer on how much AI belongs in a classroom, or on whose terms.
Common questions
Is AI being used to teach children directly?
Not in these sessions. The current training targets teachers, giving them tools to understand AI and manage how students interact with it, not to deploy AI as a teaching tool itself.
Can AI really write a student's homework for them?
Yes. Modern AI tools can produce essays and answer comprehension questions in seconds. That's precisely why many teachers want training in how to detect and discourage it.
