American schools write their rules on AI without knowing what they are preparing for

Thirty-four states and Puerto Rico have published guidance on AI in schools, and more than 70 texts have been tabled in 2026 — a regulatory frenzy that precedes any reflection on the targeted skills.

In a few years, the supervision of artificial intelligence in American schools has shifted from non-existence to saturation: 34 states as well as Puerto Rico have now published guidelines for primary and secondary education, and more than 70 legislative proposals have been submitted in 27 states in the year 2026 alone.

An article published on 7 August 2026 by eSchool News questions the logic of this race. Most of these texts deal with the prohibition — which pupils should not do with a generic model — while the preliminary question, the question of the skills that the labour market will really require, remains largely unanswered.

The precedent of computer networks

The argument put forward by Scott Laband is that educational systems repeat an error already committed: having massively oriented students towards computer programs relying on short-term signals, before code automation beats cards. Rather than stacking restrictions, the article argues to first identify the human skills that provide students with flexibility in a work environment that has become unpredictable.