New York City’s recent decision to pause student-facing generative AI tools for students in grades 2K through 8 is the right move. Young students need time to develop foundational skills, relationships, curiosity, attention, and confidence without being prematurely routed through systems that can answer for them before they learn to think through the work themselves. We have already seen what happens when powerful technologies enter young people’s lives faster than families, educators, researchers, and policymakers can build the right guardrails: social media, smartphones, algorithmic feeds. Schools should not repeat that pattern with AI.
The city’s approach is also more nuanced than a simple ban. AI literacy is not being paused. HS students will still receive lessons on AI literacy, bias, ethics, and critical thinking. A limited number of HS classrooms will participate in monitored pilots. Teachers will still be allowed to use approved AI tools for planning and administrative work. The city's position is not that schools should ignore AI, but that schools should not rush any student-facing generative AI into younger classrooms before they understand what it means developmentally, pedagogically, and relationally. In many ways, I agree with this instinct.
But an important distinction must not get lost: a pause is only useful if we are clear about what it is creating room for. This is where the next phase matters.
Right now, much of the AI-in-schools conversation still sounds like compliance. Most schools still don’t have a clear, adoptable judgment infrastructure for AI. Many of the questions being explored now are: What tools are allowed? What data can be used? Can AI grade assignments? Which products meet safety standards? Where does a tool sit in the red-yellow-green-light framework?
To be clear, these questions matter. Schools need rules and protection. Teachers need clarity, and districts need procurement standards for AI tools. But compliance isn’t the same as judgment. A red-yellow-green-light framework can tell people what type of tools are permitted, but it can’t tell a student whether using AI helps them think or avoid thinking, a teacher how to preserve the relational aspect of learning alongside AI, or even a school whether judgment-based skills are being protected and developed alongside AI. That is the missing layer.
At FutureSkills, we call this judgment infrastructure. It’s the practical layer that makes responsible AI use visible, teachable, and accountable.
It includes shared language, developmentally apt classroom norms, student-facing reflection routines, teacher-facing implementation tools, clear distinctions between support, substitution, and dependency, and structures for asking students why they used AI, what they chose not to outsource, what they changed, what they questioned, and what responsibility they still carry. This is the layer schools will need whether they pause tools, pilot new ones, or prepare to reintroduce them later.
The recent announcements from NYC and LAUSD suggest that large school systems are beginning to take AI’s risks more seriously, and that is encouraging. MIT’s recent report on AI use in teaching, learning, and research training points in a similar direction: institutions need clear policies and AI-aware educational practices that center people, community, and the learning environment.
The UK’s work on AI tutoring and sovereign education benchmarks offers another important signal. The most interesting part is that public infrastructure, curriculum alignment, safety benchmarks, and classroom evidence are being treated as part of the same problem. That is the level of seriousness this moment requires. Rules, tools, and lessons alone will not be enough. Schools need structures that help educators protect what is most human about learning while preparing students for a world where AI will be unavoidable.
For HS students, this means supporting real-world AI use with clearer expectations around authorship, accountability, bias, and discernment. For younger students, it may mean beginning without direct tool access and focusing instead on the underlying capacities AI will eventually test: attention, reasoning, patience with difficulty, source evaluation, voice, and confidence in one’s own thinking.
That’s why I see NYC’s pause as an opening. It gives school districts a year to study what younger learners actually need, better support teachers, define what age-appropriate AI readiness should look like, and build structures that make judgment easier to teach before products become embedded in the curriculum.
The equity question also matters here. Well-resourced students will continue to encounter AI with more adult guidance, more extracurricular support, and more informal access to expertise. Students in under-resourced schools are more likely to experience either restriction without preparation or adoption without protection. Neither outcome is good enough. If districts pause student-facing tools, they also need to build the structures that prepare students for future use. If they pilot tools, they need to measure much more than usage. If they approve products, they need to ask whether those tools are actually strengthening student thinking, voice, and agency. If they train teachers, they need to help them maintain the relational aspects of teaching alongside AI.
A moratorium buys time. The responsibility now is to use that time well, and at FutureSkills, we believe the next phase of AI in education should be led by judgment development and protection, not fear, novelty, or procurement cycles.
That is the work we are building toward. I’d love to hear from educators, school leaders, parents, and policymakers thinking through the same question:
What would it take to build AI policies that protect students while also preparing them to think clearly in a world already being reshaped by AI?