New York City and Los Angeles—the two largest school districts in the country—each announced last week that they are banning student use of generative artificial intelligence.
New York’s ban applies to elementary and middle schools, while Los Angeles Unified announced a moratorium for all students. Both policies grew out of parent concern over ed-tech use in schools.
Is what these districts propose—nixing a cutting-edge technology that’s quickly becoming integrated into a wide range of digital tools for most students—even possible, technically speaking?
It’s very difficult—maybe even impossible—to turn off generative AI for entire schools as if it were a light switch, said Victor Lee, an associate professor in the graduate school of education at Stanford University, who has studied the impact of AI on education.
Generative AI, the type of technology that powers large language models like ChatGPT, Claude, and Gemini, can create something new—an essay, image, a list of suggestions—in response to a question or prompt, sometimes with uncannily humanlike proficiency.
But the technology is arguably just a supercharged way of doing what an older form of AI—the predictive AI that completes text messages—would do. Both make predictions, but generative AI is different because it makes many more really fast predictions for what the next words (or “tokens” in techspeak) might be.
It’s not exactly an easy proposition to separate the two, Lee said.
Generative AI is “entering into so much stuff,” Lee said. “It’s hard to disentangle.”
For instance, the predictive text that may pop up when you’re writing a text or email message didn’t used to incorporate generative AI, Lee said. But now some models do, in fact, employ that version of the technology.
But AI bans are still worth pursuing despite the potential challenges, argued Kelly Clancy, a New York City public school mom and the leader of Parents for AI Caution in Educational Spaces, a grassroots organization.
“Saying AI is in everything, so you can’t ban it, so we shouldn’t try, is a lazy argument,” Clancy said.
Preventing kids from seeing or using AI is now harder than ever
Some forms of generative AI may be all but unavoidable, Lee and other experts said.
For instance, some of the features in a popular graphics tool, such as Canva or PowerPoint, use generative AI to help students with visuals or layouts for presentations.
Generative AI is also used in some platforms that make learning accessible for English learners (such as translation applications) or applications that read text aloud to students.
And Google Search now summarizes results using generative AI as a default and offers an “AI mode.” (Both features were rated “unacceptable” for children by Common Sense Media, a nonprofit that focuses on youth and technology. The review sited the tool’s inaccurate responses and failure to respond to red flags for unsafe behavior. )
The Los Angeles district tried to solve the problem by directing students not to interact with the AI-generated summary.
But that’s an imperfect solution, since it’s hard for students to “unsee” AI’s answers. And informational websites students may turn to for research are likely to include their own AI-generative content, which may or not be labeled as being developed or informed by AI.
“These things are just going to get harder and harder,” Lee said.
When school districts say they’re “banning” generative AI for certain groups of kids, they likely mean that they’re turning off chatbots or tools that will write a student’s essay or solve a math problem for them, Lee said.
Lee understands the two school districts’ instinct to bring more scrutiny to AI tools, particularly those that are student-facing.
But, he added, there are likely to be “weird, unexpected edge cases where a teacher or a student or a parent raises a question as to whether something can be used or not because it seems [to have] generative AI components, but it’s not the sort of thing that was initially singled out” in the overall AI use policy.
Some experts say banning AI can push tech companies to make meaningful changes
Restricting AI can push tech companies to remove some of the most problematic features from their products, Clancy said.
Similarly, Meg Leta Jones, a professor at Georgetown University who studies the impact of technology on families, thinks that districts can “turn off” the most problematic, student-facing generative AI components of ed-tech tools.
And in her mind, there’s a good argument for doing so, though she prefers school or community-level policies to sweeping, districtwide, one-size-fits-all mandates, especially in big districts like New York and Los Angeles.
AI’s impact on students is largely untested, and generative AI relies on data to power its operations—a red flag when it comes to student privacy, Jones pointed out.
“I am at least hopeful that a moratorium [like the ones New York and LAUSD have put in place] gets us to a place where we have some ideas about how we might build safe products for schools,” she said. “Right now, we’re just not even close.”
Jones, though, believes that big districts should use common sense and caution when making decisions about AI or technology in general.
“Larger districts that are covering a lot of different cultural, socioeconomic needs [should set] tech policies that represent [their] communities,” Jones said. Often, views on tech use are “not consistent across districts.”
