New York City public schools restricted access to ChatGPT on school networks and devices, citing concerns about student learning and the safety and accuracy of its output. At the same time, OpenAI said it was developing ways to help people identify text generated by the chatbot, as schools weigh the risks and possible educational uses of generative AI.
A restriction that schools can revisit
The New York City Department of Education said the restriction followed requests from schools. Its spokesperson argued that quick answers from ChatGPT do not build the critical-thinking and problem-solving skills students need for academic and lifelong success.
The measure uses the same kind of filter the school system applies to other apps and websites on school property, including YouTube and Facebook. It is not a systemwide prohibition without exceptions: individual schools can ask for ChatGPT to be unblocked.
The department also said it would welcome a conversation with OpenAI about adapting the tool for education. That leaves room for schools to consider how the chatbot might be used, even as access is restricted by default on school networks.
OpenAI says it is working on identification tools
OpenAI told TechCrunch it was developing “mitigations” to help anyone spot text generated by ChatGPT. The company described the chatbot as a research preview intended to help it learn from real-world use, and said it was incorporating feedback and lessons.
The statement also reaffirmed OpenAI’s call for transparency when people use AI-generated text. The company said its policies require users to be upfront with their audience when using its API and creative tools, and that it looked forward to working with educators on useful solutions.
Identifying AI-written work could give teachers another way to discuss how students use these systems. But the source does not describe how the proposed tools would work or establish that they can reliably determine whether a particular passage came from ChatGPT.
Why schools are concerned
ChatGPT can produce answers on subjects from poetry to coding, but its responses can sound convincing while being factually wrong. That gap between confident presentation and accuracy matters in schoolwork, where students need to assess information rather than simply accept a fluent answer.
The debate also centers on cheating and plagiarism. The article describes educators, journalists and students trying the tool on homework and essays. It cites a Wall Street Journal columnist who used ChatGPT to write a passing AP English essay, a Forbes staffer who used it to finish two college essays in 20 minutes, and an Arizona State University professor who said an essay produced for an assignment would have earned “a good grade.”
There are concerns beyond whether a student submits AI-written work as their own. The article says ChatGPT, like other text-generating systems, can reproduce public data without citing sources. It can also produce factual inaccuracies and biased material, including racist and sexist perspectives. OpenAI has introduced filters and other techniques to limit problematic output, though the source says workarounds continue to appear.
Some educators see a role for AI
Not every educator views ChatGPT only as a shortcut. Victor Lee, an associate professor of education at Stanford, suggested that it could help students discover and clarify ideas. He also described ways teachers might use the tool to generate examples that preserve the same basic content while changing style, syntax or grammar.
That kind of use could support repeated reading, reflection and revision, while giving teachers more examples to discuss. Lee also acknowledged the tension: teachers may be less pleased when a computer does much of a student’s work, and students still need to learn to write. The arrival of generative AI, in his view, raises questions about what writing instruction should teach and how it should work.
New York City’s decision puts those questions in front of school districts elsewhere. The article described the policy as apparently the first of its kind in the country and noted that ChatGPT had over a million users as of December. As researchers and companies pilot detection tools, educators face a choice about whether to restrict these systems, use them as aids, or develop ways to do both. The source leaves that question open: a filter can limit access on school devices, but it cannot settle what students should learn about AI-generated text.