Schools open the year with graded rules for AI

As the American school year opens, districts are converging on a common shape for AI rules: access graded by age. Katy Independent School District in Texas bars students up to sixth grade from open-ended chatbots such as Gemini and Copilot, allowing only curriculum-embedded tools; seventh graders get district-approved tools under supervision; high schoolers may use generative AI for specific purposes with teacher authorisation. Vermont's state framework follows the same curve, from no chatbots before third grade to broad AI fluency work in high school, and Illinois issued state guidance for districts in July.

The pattern is recognisable — it is how schools eventually handled calculators and internet access — but compressed from a decade of argument into about two years. The tiering embodies a defensible theory: that the risk of an open-ended assistant is inversely proportional to a student's ability to evaluate what it tells them, and that the skill of using one well is worth teaching before adulthood.

The numbers explain the urgency and the futility in equal measure. Surveys cited in coverage of the new policies report that 86 percent of nine-to-seventeen-year-olds have used generative AI, about a quarter of them daily, and that 85 percent of young users apply it to schoolwork. The tool arrived before the rules, and the rules govern school accounts and school networks — not the phone at the kitchen table where most of the schoolwork use actually happens. A district policy can shape what school-managed AI looks like; it cannot make eighth-grade essay writing AI-free, and policies written as if they could will be quietly ignored.

The gap that deserves more attention is on the other side of the desk. The share of teachers reporting no AI training at all has fallen from 60 percent in late 2024 to 42 percent — meaning that two school years into the era, four in ten teachers are still expected to referee a technology nobody has shown them how to use.