Two significant things happened on May 27th. Let’s get into it.
What Happened on May 27
On May 27, Randi Weingarten, president of the American Federation of Teachers, gave a nearly hour-long speech at the National Press Club calling for a ban on student-facing AI in elementary schools, no screens in classrooms through second grade, and a new tax on big tech companies to fund teacher training. She framed her 10-point plan under the slogan “Devices down, eyes up, hands-on.”
The same day, New York City Schools Chancellor Kamar Samuels told an audience at Bank Street College that the city’s draft AI guidance—released in March—had “missed the mark.” The city had collected more than 6,000 comments on that guidance over 45 days: 48% from educators, 42% from parents. What Samuels had initially read as public fear turned out to be something sharper—anger, and “a lack of trust in institutions, a lack of trust in our security mechanisms, and also a lack of trust in or deep skepticism of ed-tech companies.”
That backlash had been building for months. In May, parents spent seven hours at a New York City school board meeting demanding a two-year AI moratorium in city schools. A petition for the moratorium gathered more than 3,000 signatures. Elected officials called on the Education Department to pause its AI guidance entirely.
Two major institutions, on the same day, moving toward more restriction. That’s not a coincidence.
The Line Weingarten Drew
Weingarten’s position looks contradictory at first. She partnered with OpenAI and Microsoft in July 2025 to provide AI training to AFT’s 1.8 million members. She wants teachers using AI. The day before her May 27 speech, she visited Newark schools and watched elementary students use Khanmigo, Khan Academy’s AI tutoring chatbot. The students praised it. Then she went to Washington and called for banning student-facing AI in elementary schools.
A lot of people read that as inconsistency. She drew the right line.
There’s a real difference between AI tools that help teachers do their jobs and AI tools that do the intellectual work students are supposed to be doing. Weingarten drew that line explicitly, using a specific term: cognitive offloading. Students outsourcing the thinking that school is designed to build. She described it as one of the central dangers of the current moment—students relying on AI for tasks “they’d otherwise do in school,” gradually reducing their ability to do that work independently.
She’s right that cognitive offloading is the real problem. The question is whether restricting what students can access is the right response to it.
The Bronx Classroom
At the same Bank Street event, Shael Polakow-Suransky—former deputy chancellor of NYC schools, now president of Bank Street College—described visiting a Bronx classroom where students were working on translating fractions into decimals using an AI tutoring program. Most of them eventually got to the right answer.
Almost none of them showed any conceptual understanding of what they were doing.
The teacher was so occupied troubleshooting the software that he couldn’t spend time with any of his students.
That scene captures the real problem better than most policy proposals manage to. The technology worked as designed. Students produced correct outputs. The teacher was in the room. And nobody learned how fractions work.
Cognitive offloading isn’t a future risk. It’s already happening in classrooms that look, on the surface, like they’re using technology effectively.
What the Debate Should Be About
The instinct to restrict AI for students while keeping it available for teachers makes sense once you understand what it’s protecting. A teacher using AI to draft lesson options, generate differentiation suggestions, or analyze patterns across assessment data is offloading administrative and logistical work—the kind of task that shouldn’t require a professional’s full intellectual bandwidth in the first place. A student using AI to write an essay is offloading the intellectual work the assignment was designed to build.
But a blanket age-based ban on student-facing AI is still a blunt instrument.
A student using AI to pressure-test their thesis against counterarguments might deepen their thinking. A student using AI to surface a list of historical examples they then have to analyze, contextualize, and connect to an argument themselves might be doing more intellectual work than they would have without it. A student pasting a prompt into ChatGPT and submitting whatever comes back is the Bronx fraction problem in an essay format.
The difference isn’t whether the tool is present. The difference is whether the student is doing the thinking.
That’s a harder policy to write than a blanket ban. It requires teachers to understand what cognitive offloading looks like in their specific classroom—which is different in an AP World History seminar than in a second-grade reading lesson. It requires ongoing professional judgment, not a one-time decision from a district office.
What This Means Right Now
Weingarten’s plan will shape policy debates through the fall. New York City’s final AI guidance will be tighter than the March draft. Ohio requires all school districts to have formal AI policies in place by July 1. Across 31 states, 134 bills related to AI in education are under active consideration.
Most of those policies will draw lines based on grade level or approved tool lists. That’s not wrong. But it doesn’t resolve the central question for the teacher in the room: whether students are actually doing the thinking.
That question isn’t answered by what’s blocked on a school network. Students in secondary school have access to every AI tool on their phones and home computers. District policy covers school devices during school hours. It doesn’t touch the larger reality of how students actually work.
The only lever teachers genuinely control is the assignment itself—whether it requires evidence of thinking that can’t be outsourced. In-class writing. Oral explanation. Process artifacts that show the work happening, not just the final product. None of this is new pedagogy. AI has just made the need harder to ignore.
Takeaway for Teachers
This week, take one major assignment you’re planning in the next month and ask one question: can a student produce the final product without doing the thinking this assignment is supposed to build? If yes, add one element that makes the process visible—a short reflection on how the argument changed, a verbal explanation of a key decision during class, an in-class draft checkpoint. You don’t have to redesign the whole assignment. Just make the thinking observable somewhere.
The goal isn’t to pretend AI doesn’t exist. It’s to make sure the intellectual work still happens—in your classroom, with your students, visibly enough that you can see it.
If you’re working through how to design assignments that hold up in the AI era—and how to explain that reasoning to students and parents—my book goes deeper. The AI Doesn’t Know Your Students is available on Amazon and at shouldiuse.ai/book.
David Jacobson is a high school history teacher. He writes about AI, education, and the messy intersection of the two at shouldiuse.ai.
