Summary of Key Points
New York City has implemented a one-year “pause on generative AI for students” policy for grades 2-K to 8 (approximately 600,000 public school students), while also restricting screen time for younger children. This is not an anti-AI measure; rather, it aims to reframe the relationship between AI and children based on age: younger students are temporarily barred from using generative AI directly, while high school students can use it in limited amounts under supervision. Teachers can still use AI for lesson preparation, with exceptions for students with special needs. Many other countries around the world (such as France, Japan, and China) have adopted similar tiered approaches. The core debate revolves around whether AI should serve as a “scaffold” to aid in cognitive development or a “task executor” that replaces independent thinking. Ultimately, the question is whether education should first focus on cultivating fundamental abilities that allow children to function without relying on tools, before teaching them how to use AI.
I. New York’s “Pause” is Not a “Ban,” but a “Age-Based Boundary”
New York’s policy establishes an age threshold for the use of AI:
- Who is affected? Students in grades 2-K to 8 (approximately 6–14 years old) are not allowed to use generative AI designed for students, and the AI features previously approved in 38 educational programs will also be disabled.
- Who is exempt? Teachers can use AI for lesson preparation and administrative tasks; students with disabilities and those learning multiple languages (e.g., those who are not fluent in English) can continue to use AI for necessary assistance. High school students can still access AI, but they must learn critical thinking skills twice a year and use approved tools under the supervision of a teacher, with time limits in place.
- Why this approach? New York aims to give younger children the opportunity to think for themselves. For example, when stuck on a math problem, they are encouraged to spend more time trying to solve it on their own rather than immediately seeking answers from AI. Similarly, taking a moment to ponder while writing a composition is seen as beneficial for cognitive development.
II. A Global Trend of Setting Age Limits for AI Use
This is not a coincidence; educational policies around the world are being more cautious with younger students and gradually allowing older students to use AI:
- UNESCO recommended in 2023 that students under 13 years old should not use generative AI independently.
- France has more specific guidelines: elementary school students can only learn about AI without operating it; middle school students (around 14 years old) can use it with teacher supervision; high school students can use it independently within certain limits, but copying AI-generated work is considered cheating.
- China’s 2025 guidelines explicitly state that elementary school students are prohibited from generating content independently, middle school students can analyze the logic of AI-generated content, and high school students can explore the technical principles behind it.
The common theme of these policies is a concern that AI should not complete the most challenging parts of the learning process before children have developed the ability to think for themselves. For example, if AI provides the solution to a math problem before the child understands the concept, they may never learn how to solve it independently.
III. AI as a “Scaffold” or a “Task Executor”? The Difference Lies in the “Thinking Process”
There are two fundamental ways to use AI in education, with vastly different outcomes:
- As a scaffold: For example, students at Renmin University High School use AI to cluster historical data or simulate population growth. AI helps process large amounts of data or conduct complex experiments, while teachers guide them in analyzing the results and understanding the underlying logic. This approach uses AI to support learning.
- As a task executor: For example, students simply submit questions to AI for answers or ask AI to write draft essays. In this case, AI performs the core thinking processes (formulating ideas and organizing language) on their behalf.
Why is the latter approach problematic? In learning, “productive failure” (e.g., making mistakes and correcting them) is essential for deep understanding. AI simplifies this process, similar to taking the elevator instead of climbing stairs, which may lead to a lack of practical skills.
IV. A Strict Ban on AI Could Create New Inequalities
A blanket ban on AI could lead to two issues:
- Wider wealth gaps: Children from wealthy families can continue to use AI at home or hire tutors to teach them how to use it properly, while those from poorer families may not have access to such resources, putting them at a disadvantage in the workplace.
- Easy avoidance of bans: If assignments are graded based on “standard answers,” students may use AI to quickly complete them, as “speed” is often more valued in the current evaluation system.
The real solution is to provide guidance, such as using “process-based evaluations” (evaluating drafts, classroom discussions, and oral explanations) rather than solely focusing on the final results. Technology companies can also develop AI that asks questions (e.g., “What step have you considered so far?”) before providing answers.
V. Don’t Impose Adult Efficiency Anxiety on Children
The adult world is obsessed with AI efficiency—companies train employees to use it, and programmers use it to write code. However, children are not “machines to be optimized,” and childhood is not a “production line for training.” A 10-year-old child spending 40 minutes on a problem may seem inefficient, but during this process, they learn to persevere, identify their own mistakes, and complete tasks independently. These skills are more valuable than 100 correct answers generated by AI.
In the AI era, we should ask whether we can teach children to think independently before relying on AI. Instead of making them accustomed to seeking AI’s help from a young age, we should focus on developing their ability to think for themselves.
Conclusion
New York’s temporary policy is not the perfect solution, but it highlights the essence of education: to cultivate well-rounded individuals, not just tools that can use AI. We need to teach children how to use technology while ensuring they develop the ability to think, express themselves, and make independent judgments. In the future world, while AI can provide answers, it is the children who must learn why questions are asked, how to evaluate them, and how to apply them to real problems. Let’s not rush children through a stage of awkward growth due to adult concerns about efficiency. It’s okay to take things slower.