By Better Together Agency
NYC Public Schools is right to protect foundational learning and put guardrails around student-facing AI. A restriction on use should not become a delay in AI literacy.
For the 2026 to 2027 school year, New York City Public Schools is implementing a one-year moratorium on student-facing generative AI in grades 2-K through 8. High school students will have limited access to approved programs, and every student in grades 9 through 12 will complete two 45-minute AI literacy modules covering how AI works, data privacy, bias and appropriate school use. The district also has separate screen time guidance by grade. NYC Public Schools outlines the policy here.
The concerns behind the policy deserve to be taken seriously. Students need time to read, write, create, reason, solve problems and work through difficulty without handing the work to a machine. Younger children do not need unrestricted access to chatbots. Schools also have legitimate questions about privacy, bias, inaccurate outputs and how generative AI may affect learning.
Where Better Together Agency disagrees is with the idea that protecting those skills requires waiting until high school to make AI literacy a formal part of a student’s education.
This is the strongest argument for a cautious approach, and there is evidence behind the concern.
Brookings has warned about cognitive offloading when students use generative AI to avoid effortful learning. Its 2026 work on AI and education stresses the importance of productive struggle, critical thinking, writing and analysis. At the same time, Brookings also says narrow, targeted AI use can support learning when it is designed around sound teaching. Read the Brookings analysis.
That distinction matters. Students should be expected to think without AI. They should also learn how to judge AI when they encounter it.
A middle school student does not need a chatbot to write an essay for them. The same student can learn why an AI-generated answer may be wrong, how to check a source, why a system may reproduce bias and when using AI would interfere with the learning goal.
Those lessons protect thinking rather than replace it.
Concerns about excessive screen use are real. They should not be used as a catchall answer to AI education.
The American Academy of Pediatrics updated its guidance in 2026 to describe a broader digital ecosystem rather than treating every digital experience as equivalent screen time. Its policy distinguishes between technology designed around children’s developmental and learning needs and systems designed to maximize attention, engagement and data collection. Read the AAP policy statement.
Teaching a student to recognize an AI-generated falsehood does not require hours on a screen. Neither does a classroom discussion about privacy, bias or why a human should verify an AI answer.
Schools can reduce unnecessary screen use and still teach AI literacy.
One reason this debate gets stuck is that AI education is often treated as if it means teaching children how to get better answers from a chatbot.
AI literacy is broader. UNESCO’s AI Competency Framework for Students includes a human-centered mindset, ethics of AI, AI techniques and applications and AI system design. It calls for AI learning objectives to be integrated into school curricula so students can engage with AI safely and responsibly. See UNESCO’s framework.
For younger students, age-appropriate AI literacy could include questions such as:
None of this requires unrestricted access to generative AI.
Another argument is that children will catch up quickly. Chatbots are easy to use. Four years of high school is plenty of time to learn the tools.
There is truth in part of this. Today’s interfaces may be irrelevant by the time a sixth grader enters the workforce. Teaching a specific product should not be the goal.
The harder skills are judgment, verification, disclosure, privacy awareness and knowing when AI should not be used. Those habits are not tied to one product. They can develop over time, much like media literacy, research skills and digital citizenship.
Waiting to teach those habits assumes students will wait to encounter AI. Many will not.
People who support tighter AI restrictions often point to social media. Society gave children access to technology before fully understanding its effects, and schools and families are still dealing with the consequences.
That history supports caution.
It also supports earlier education.
Young people used social platforms before many adults taught them how algorithms influence what they see, how companies collect their data or how misinformation spreads. The lesson should include safeguards, adult responsibility and literacy from the start.
AI brings different risks, but the governance principle carries over. Access without education is a poor model. So is assuming a technology disappears from a child’s life because a school restricts it.
This may be the hardest implementation problem.
Many teachers have received little preparation for teaching responsible AI use. Adding another requirement without time, training and clear standards would be unfair to educators and difficult to execute well.
That is an institutional readiness problem. It calls for teacher education, approved tools, privacy rules, clear boundaries and assignments designed for the presence of AI.
Schools should not place the burden on individual teachers to figure out AI governance on their own.
Better Together Agency is not proposing a national curriculum or claiming to know the right developmental standard for every child. Our work is in responsible AI, bias, governance and communications. From that perspective, a phased model deserves serious consideration.
Focus on teacher-led literacy rather than independent chatbot use. Students can learn that AI systems are created by people, can make mistakes, can reflect bias and should not receive private information.
Add controlled demonstrations and source-checking exercises. Students can compare an AI answer with trusted sources, identify unsupported claims and discuss when AI use would undermine the purpose of an assignment.
Expand guided use. Require disclosure when AI is used. Build assignments where students must explain, defend and verify their work. Teach bias, privacy, intellectual property, appropriate use and the limits of automated systems.
Protect time for learning without AI. Train educators. Set clear rules for data and approved tools. Give families understandable information about what students are using and why.
The debate does not require choosing between unrestricted student access and keeping generative AI out of K-8 classrooms.
Schools have more options.
Students need opportunities to develop reading, writing, creativity, problem solving and independent judgment without AI doing the work for them. They also need adults who can teach them how AI works, where it fails, how bias enters systems, what information to protect and when human judgment needs to take over.
Those goals can exist together.
New York City’s one-year moratorium creates time to study the effects of generative AI on younger students. That time should also be used to ask a harder question: What would age-appropriate AI literacy look like before high school?
For the 2026 to 2027 school year, NYC Public Schools has a one-year moratorium on student-facing generative AI in grades 2-K through 8. The restriction applies to student-facing generative AI, not every form of technology or AI used by schools.
Yes. Students in grades 9 through 12 are required to complete two 45-minute AI literacy modules on how AI works, data privacy, bias and appropriate school use. Some students will have additional approved AI use through pilots and career programs.
Yes. AI literacy can include source verification, privacy, bias awareness, understanding AI errors and learning when not to use AI. Many of these concepts can be taught through teacher-led lessons and controlled exercises.
Students are already encountering AI in search, apps, media and other digital systems. Earlier literacy can help them question what they see, protect personal information and recognize when an AI system should not be trusted.
Better Together Agency works with organizations on responsible AI, bias, governance and communications. Read our Biases in Generative AI research.