Big Tech Wants AI in the Classroom. These Teachers are Pushing Back.

Despite all the hype surrounding the release of ChatGPT in November 2022, I only started using large language models on a regular basis last month.

It's not that I'm reflexively opposed to new technologies. I've worked at tech startups. I research and write about tech policy. I'm a millennial, after all – we came of age with the internet. But I value routine and tend to stick with what works. I had a flip phone until I graduated college. I rarely use social media. I still don't see the value-add of virtual assistants like Alexa or Siri.

It takes a true "aha moment" for me to adopt a new technology. And that's precisely what I experienced recently when I spent a weekend trying out paid versions of Anthropic's Claude and OpenAI's GPT 5.2.

Remake my website in the image of The Atlantic? Done. Show me five job postings that best match my background. Voila. Any ideas for blog post topics? Here are three, with outlines and narrative arcs to boot.

I was hooked. Think of the productivity gains. Think of the time savings. And yes – I know I was late to the game.

After a few days riding high, I started to grapple with more uncomfortable questions: What if I become so reliant on these tools that I lose the ability to form my own thoughts or generate original insights? When is it appropriate to use AI, and when does it cross an ethical line? If anyone can now produce a (seemingly) flawless policy memo, what's the point of striving to write one at all?

If I'm struggling with these questions, I wondered, how is the next generation navigating this new terrain?

To find out, I interviewed six high school teachers and dug through the latest research on how generative AI is being used in schools today. Here are the most striking takeaways.

When I asked Eric Colburn, an English teacher at Brookline High School, how many students at BHS are using AI, he didn't hesitate: "Close to 100%".

"Of course, it's hard to know for sure," he added with a smile.

Mr. Colburn's estimate sounded high – or maybe I just couldn't believe all these 16-year-olds were more tech savvy than me. But the research suggests he's not far off.

The Cypress, Brookline High School's student news site, surveyed the student body last November and found that 3 in 4 students reported using AI for schoolwork. That same month, Gemini and Notebook LM were made available to all students for free through their Google Suite, meaning usage rates have likely climbed since then.

"By the end of last year, I realized my students couldn't do an assignment without AI now. They would fail if I told them to do it without AI," Peter Sedlack, an English teacher at Brookline High School, told me. "I think some of them, within the year, had lost the ability to actually do the work."

Nationally, school districts have done a sharp about-face on AI since 2023, with adoption rates surging in 2025.

New York City was among the first to ban the technology outright in early 2023, citing concerns over cheating and harm to student learning. By the end of that year, the district had reversed course, lifting the ban and launching an AI Policy Lab to guide its approach to integrating the technology. In 2025, Miami-Dade County Public Schools, the nation's third-largest school system, rolled out Google's Gemini for more than 100,000 high school students. Broward County, the nation's sixth-largest district, introduced Microsoft's Copilot for thousands of teachers. Many other districts have followed, experimenting with custom AI platforms and personal assistants for teachers and students, with mixed results.

The Center for Democracy and Technology found that over 85% of high school teachers and students have now used AI for schoolwork or personal use. A Pew Research survey released last month put the figure of teens using AI chatbots specifically for schoolwork at 57%. Notably, parents in higher-income districts report higher rates of AI use for themselves and their children.

With most districts effectively giving students the green light, many haven't hesitated to use it – often to cheat.

"It was pretty distorting to have students who I'd built relationships with and not being able to look them in the eye and trust their work anymore," Mr. Sedlack told me. "You would read an essay and immediately realize this probably wasn't somebody's work. And then spending an hour of my time trying to hunt down evidence — it isn't emotionally sustainable."

In the early 2010s, Google began offering Chromebooks to K-12 schools at steeply discounted rates. The pitch was hard for schools to refuse: Cheap laptops that plugged seamlessly into tools districts were already using. The market did the rest. By 2025, Chromebooks held 93% of the K-12 market.

A similar strategy is playing out today. OpenAI, Google, and Microsoft have each built education-centered versions of their AI products – ChatGPT for Teachers, Gemini for Education, Copilot – and are competing aggressively for their slice of the $252 billion ed-tech industry, which is projected to double in the next five years. Microsoft and Google alone have pledged over $5 billion to bring AI into K-12 education.

Notably, the investments come wrapped in a morale rationale, easy to spot in any corporate press release on the subject: students who aren't trained on AI today will be left behind in tomorrow's job market.

Perhaps the biggest splash came last July, when Microsoft, OpenAI, and Anthropic announced a $23 million partnership with the American Federation of Teachers (AFT) to launch the National Academy of AI Instruction, which aims to train 400,000 teachers — a tenth of the U.S. total — in AI over the next five years.

Partnering with the country's largest teachers union gave the tech companies a veneer of grassroots buy-in. But many teachers aren't buying it.

In an email to BHS faculty, Mr. Dickerman compared the arrival of AI to a predator in a henhouse: "Are there foxes in the woods out there? Sure, but that's why we build hen houses: to protect those within. Now some unknown person or group has unilaterally decided to bring the fox inside. We should do everything we can to prevent that from happening here."

AFT President Randi Weingarten defended the partnership, arguing it was necessary to establish common-sense guardrails in the absence of federal regulation, and to ensure teachers remain in control of classroom technology. Meanwhile, tech companies insist they want teacher input. "To best serve students, we must ensure teachers have a strong voice in the development and use of AI," said Brad Smith, Microsoft's vice chair and president.

But the teachers I spoke with aren't convinced. "Most teachers are skeptical at best," English teacher Ben Berman told me. "It's too disruptive to most people's practices."

Interestingly, it was OpenAI's chief global affairs officer Chris Lehane, speaking after the AFT launch in July, who put the stakes most plainly: "When it comes to AI in schools, the question is whether it is being used to disrupt education for the benefit of students and teachers or at their expense."

That's a striking admission from someone who's company is aggressively pushing its product into classrooms nationwide.

Who is in charge of AI policy for public schools in the United States?

It's a seemingly simple question, and one I put to every teacher I spoke with. But the answer is murky.

At both the federal and state level, guidelines and best practices have superseded legislation. President Trump signed an executive order last April calling on schools to integrate AI into all subject areas, and in June announced that more than 60 companies had pledged to make AI products and training available to schools. Thirty states have published guidance on generative AI use in schools, and a handful have launched pilot programs. But concrete legislation is nearly nonexistent – most states have simply punted the hard questions to individual districts.

In Massachusetts, only 9 of 300 school districts have adopted AI policies. Nationally, only 31% of U.S. public schools have an AI policy, and only 18% of principals report that their school or district has provided any guidance on AI.

The vacuum exists for a reason. At the district level, responsibility for AI policy is diffuse – spread across school boards, principals, superintendents, and labor unions, with no clear owner. Turnover in administrative offices makes sustained leadership on the issue even harder to maintain.

"There has not been a district-wide initiative," Mr. Berman told me of Brookline High School. "That's partially due to turnover in the central office — we've had tremendous turnover. And partially because people's feelings around AI and its usage are so varied."

"Schools are playing catch up," said Elizabeth Laird, director of equity and civic technology at Center for Democracy and Technology. "The adoption of this technology, in many cases, is ahead of the policy."

Into that vacuum, the companies with the deepest pockets and the most to gain have stepped in as de facto policymakers.

When most people think about AI today, they picture the large language models – ChatGPT, Claude, Gemini, Grok – and the companies behind them. The teachers I spoke with were nearly unanimous on this point: these tools have no place in the student learning experience. They suppress critical thinking, and students use them to cheat.

"A big part of writing is developing your own voice and your own sense of self," Mr. Sedlack told me. "AI sanitizes that. When it tries to be creative, it's flat and meaningless — a bunch of almost-cliché aphorisms that sound smart, but only to those who aren't actually smart."

But there's an entire industry of "AI wrappers" – proprietary platforms built on top of general-purpose large language models – designed to be kid- and teacher-friendly while offering more controlled, customizable learning experiences.

Aurora Public Schools, a 38,000-student district in Colorado, reported a 28% improvement in student literacy after introducing a platform called MagicSchoolAI in 2024. Many students at BHS use Flint to generate interactive quizzes and get one-on-one tutoring. There are "Socratic bots" that refuse to give students direct answers, responding instead with questions designed to push their thinking. Notion AI can summarize notes and convert them into study guides and flashcards.

Teachers are generally more receptive to tools like these because they make it harder for students to simply outsource their thinking. Mr. Berman also sees some AI chatbots as a more affordable alternative to private tutoring: "A human tutor runs a hundred dollars an hour," he told me. "But you can get Khanmigo, which is a very good AI tutor and it won't do any work for you."

Growing up, I typically spent two to three hours a night on homework. So it took me aback when teacher after teacher told me they basically no longer assign any.

"100% of the writing students submit happens in class," Mr. Colburn told me.

The AI revolution has pushed many teachers toward a more analog style of teaching, with more in-person work, more assignments completed on paper, and less sent home. Mr. Berman put it to me like this: "If anything, since AI, the swing has been the other way, to really think about how do we return to a tech-free space. Most teachers are really shifting practices — away from long-form writing, away from giving kids access to write at home, doing a lot more in-class writing as prevention."

Mr. Sedlack described a similar shift in his own classroom. "It's forced a pedagogical shift in how I teach writing," he told me. "I've had some experiments with having students take one paragraph home to rewrite after I've given feedback, but it's not great. There's no stopping them once they go home."

Other teachers echoed this idea of the classroom as a sanctuary from tech. "I don't want them staring at screens and just asking all their questions to the computer," Melissa Nixon, a physics teacher at BHS, said. "I want them to play around and get messy, but also know how to talk to each other."

The trend extends beyond Brookline. Inside Higher Ed declared 2025 the year of the blue book comeback, as teachers across the country revived the old in-class exam booklet as a defense against ChatGPT.

Carlo Rotella, an English professor at Boston College, describes the AI-resistant classroom as one that emphasizes process over product, oral and handwritten assessment, and what happens between people in the room. Chanea Bond, an English teacher in Fort Worth, has also embraced the analog approach: "If you walk into almost any one of my classes today, you will see that all of my students are handwriting, and they are journaling, and they are constantly and consistently doing everything with a pen or a pencil."

But there are skeptics of the analog retreat. Greg Kulowiec, AI adviser to the Massachusetts Department of Elementary and Secondary Education, runs AI literacy workshops for teachers and students and argues that avoidance only makes things worse: "If I promote AI, then I can teach how to effectively use it," he says. "If I ban or ignore it, then students are going to be left to their own understanding, which might be sophisticated, and it might not."

The fundamental question is this: how do you prepare students for an AI-dominant world without short-circuiting their critical thinking? Nobody has a clear answer.

But one thing does seem clear: the companies best positioned to profit from AI in schools are the ones currently shaping how it gets introduced and used, rather than the teachers who have spent their careers in the classroom.

The equity argument deserves particular skepticism. There is a real access gap: 67% of low-poverty districts have introduced AI training for teachers, compared to only 39% of high-poverty districts. Tech boosters invoke the One Laptop per Child program of the 2000s as a precedent for closing this gap. But children's agencies like UNICEF see One Laptop per Child as a cautionary tale, not a success story: "With One Laptop per Child, there was a ton of wasted expenditure and poor learning outcomes," writes Steven Vosloo, a digital policy specialist at the organization. "Unguided use of AI systems may actively de-skill students and teachers."

Mr. Dickerman put it more bluntly. Pushing AI tools on underprivileged students, he argues, doesn't close the equity gap. It widens it in ways that are harder to see. "The future of humanity may be a world of thinkers and non-thinkers," he told me. "If you take kids from backgrounds where they don't have the same supports at home as the rich kids and you're just consigning them to the bottom, you're not closing the equity gap. You're just giving up on them."

Alex Molnar, a director of the National Education Policy Center at the University of Colorado at Boulder, imagines where this all leads: students who can't explain their own thinking, teachers who can't explain their evaluations, administrators who can't explain their decisions – and local funds flowing quietly to distant corporations. "We essentially will have transformed public education from a civic institution into a portal for funneling money to private interests."

The teachers I met are doing their part to keep this from happening. The question is whether anyone else is listening.