Should AI be banned in schools?
No, not as a standing policy. New York City, Los Angeles, and Seattle each banned ChatGPT in schools between December 2022 and January 2023, and two of the three reversed course within about a year once enforcement broke down.
That track record, not the general debate, is the most useful evidence available on whether a ban actually works. This is a narrower question than the pros-and-cons debate covered in our AI in schools pillar, and different from whether AI should be allowed with conditions, covered in should AI be allowed in schools. This article is specifically about prohibition: does blocking AI outright work, and what happened where it was tried.
The core case on each side, before the district-level detail:
- For a ban: fewer opportunities to submit AI-written work as a student's own, and a clearer signal that unsupervised assignments should reflect the student's actual thinking.
- For a ban: protects instructional time for building skills like drafting and problem-solving before shortcuts are available.
- Against a ban: most AI use happens on personal devices off school property, where a network block has no reach.
- Against a ban: students still need to learn the tools they will use in college and at work, and a ban delays that instead of guiding it.
Real bans, and what happened next
Three large districts banned ChatGPT within weeks of each other in late 2022 and early 2023. What happened afterward varied more than the initial policy did.
| District | When banned | Why | What happened next |
|---|---|---|---|
| New York City Public Schools | January 2023 | Concerns ChatGPT would not build the "critical-thinking and problem-solving skills" the district wanted, plus cheating worries | Reversed by May 2023; built its own classroom guidance instead of blocking access |
| Los Angeles Unified School District | December 2022 | Academic honesty concerns over ChatGPT on school Wi-Fi and devices | Announced its own AI chatbot, "Ed," in August 2023; soft-launched to about 55,000 students the following spring |
| Seattle Public Schools | December 2022 | Wanted "original thought and work from students," banned ChatGPT plus six other AI writing tools | Ban has held; no public reversal reported as of this writing |
New York City Public Schools banned ChatGPT on school devices and networks in January 2023. By May 2023, the district reversed the ban. Chancellor David Banks wrote that the "knee-jerk fear and risk" of the initial policy had overlooked the technology's potential and the reality that students would need to understand generative AI for the world they were entering. (Forbes, NBC News)
Los Angeles Unified School District blocked ChatGPT from school Wi-Fi and devices in December 2022. It announced a homegrown replacement, a chatbot called Ed, at a splashy event in August 2023, but the $6 million system did not actually reach students until a soft launch the following spring, when it rolled out to roughly 55,000 students across 100 schools. (EdSurge)
Seattle Public Schools banned ChatGPT on all school devices in December 2022, along with six other AI writing tools including Rytr, Jasper, and WordAI, saying the district "requires original thought and work from students." (GeekWire)
All three were large, well-resourced urban districts with real capacity to enforce a network-level block, and two of the three walked it back within roughly a year.
Why network bans struggle to hold
The core enforcement problem is simple: a school network ban only ever covered school Wi-Fi and school-issued devices. It never touched a student's phone on cellular data, a home computer, or any AI tool used after the school day ended. By December 2025, RAND's American Youth Panel found 62% of students ages 12 to 29 had used AI for homework, up from 48% just seven months earlier. Homework, by definition, mostly happens off school premises. A ban that only reaches the building doesn't reach where most of the actual use happens.
There's also a monitoring cost. Enforcing a ban means a school has to actually detect violations, and the detection tools available for that are shakier than they're marketed to be. Turnitin's AI detector is advertised at 98% accuracy, but independent testing, including a University of Chicago Booth working paper, found sentence-level false-positive rates closer to 4%, with a documented bias against non-native English speakers: seven different detectors flagged 61.3% of TOEFL essays as AI-written versus close to zero for native-English writing. A ban backed by an unreliable detector produces false accusations as a side effect, which is part of why Vanderbilt University disabled Turnitin's AI-detection feature entirely in 2023.
Where a ban does still make sense
None of this means every restriction is pointless. Age-gating independent, unsupervised use of open AI platforms is a specific recommendation in UNESCO's 2023 guidance for generative AI in education, aimed at younger students rather than a blanket policy for an entire district. A narrower restriction, say, no open chatbot use during a specific assessment, or no AI access for students below a certain grade, is a much easier thing to actually enforce than a district-wide network block, and it targets a specific risk rather than trying to prohibit the technology everywhere at once.
The alternative most districts have landed on
The pattern across NYC, LA, and the broader district landscape has been a shift from prohibition to permission with structure: sanctioned tools, disclosure requirements, and teacher training, rather than either a ban or unrestricted access. UNESCO's guidance frames this the same way, treating outright bans as a weaker tool than clear policy plus training. We lay out what that structured-permission approach actually looks like in should AI be allowed in schools, and the substantive case for why some uses genuinely shouldn't be allowed, ban or no ban, in why AI should not be used in schools.
A five-day reset, if you want one
For parents and teachers weighing this, the fix is rarely a blanket ban, it's a few deliberate habits around how AI gets used. The free 5-Day AI Reset walks it in five short emails: measure where AI shows up, take one task back, name the urge, sort helpful use from skill-replacing use, and set rules that survive a bad week. No quitting AI, no guilt.
Frequently asked questions
Have any major school districts actually banned AI?
Yes. New York City, Los Angeles, and Seattle all banned ChatGPT on school networks and devices in late 2022 or early 2023. New York City and Los Angeles both reversed course within about a year and moved toward building their own AI tools and guidance instead.
Why did New York City reverse its ChatGPT ban?
The chancellor said the initial ban reflected "knee-jerk fear" that overlooked the technology's potential and the reality that students would need to understand generative AI for the world they were entering into, both in future schooling and work.
Do AI bans actually stop students from using it?
Not effectively. Bans typically only cover school networks and devices, while most AI use for homework happens outside school hours on personal devices. Usage data shows AI use for homework has continued rising even in districts with restrictive policies.
What is a more targeted alternative to a full ban?
Age-gating independent use for younger students, or restricting AI access during specific graded assessments, targets a specific risk without trying to enforce a district-wide network block that's hard to police and easy to route around.
Is banning AI the same as arguing it shouldn't be used?
No. Banning is a specific enforcement mechanism. Whether AI use is actually harmful for certain school tasks is a separate, substantive question, which we cover directly in why AI should not be used in schools.
Why do critics say AI should not be banned in schools?
The main argument is that a ban does not restore the learning it is meant to protect. It pushes AI use underground, rewards students who already know how to route around a network block, and does nothing to prepare students for workplaces where these tools are already standard.
What are the practical downsides of a school AI ban?
Two, mainly. Detection is unreliable: independent testing found AI-detection tools flag native-English writing far less often than non-native English writing, producing false accusations as a side effect. And coverage is incomplete: a network ban only reaches school Wi-Fi and devices, so it misses the majority of AI use, which happens on personal devices after school.
This article is informational and not a substitute for school-specific policy guidance.
Part of our guide to AI in schools: the real pros and cons. See also: should AI be allowed in schools · why AI should not be used in schools.