What stuck with me, the more I sat with yesterday's Senate HELP hearing and went back through the clips, is that the real question isn't whether AI belongs in schools, or even how to use it. It's which AI.

"AI in schools" isn't one thing. A consumer chatbot built to maximize engagement — what InnovateEDU's Erin Mote calls a Yes-Bot that indulges rather than teaches — is a different tool from a FERPA-bound, purpose-built tutor, and the two carry different risks. Mote's fix is scalpel, not sledgehammer: heavier guardrails on the high-risk consumer products, lighter ones on vetted educational tools. Get the distinction wrong and you either wave through the harmful stuff or smother the useful stuff.

This isn't a DC abstraction. It's already remaking classrooms at every level. The New York Times' The Daily walked through the year AI reshaped K-12: the cheating panic, a nine-figure push from tech companies to get chatbots in front of kids, and a federal AI-education effort still short on a working definition of "AI literacy." In higher ed, The Chronicle found that 65 percent of the professors it polled had caught students cheating with AI, and nearly all of them are rebuilding how they teach in response. One philosophy professor put it the way I would: the question isn't whether students use AI, but how — and whether they're still doing the thinking.

Here's the uncomfortable part. We keep saying to judge these tools by outcomes, not hype, but there's not yet a single high-quality causal study on AI's long-term effects on learning, which is why Chairman Tuberville and Ranking Member Blunt Rochester asked the GAO to investigate. We're being asked to regulate, and to overhaul teaching, around a technology we can't yet measure.

And the policy machinery isn't waiting. A coalition including Mike Pence's Advancing American Freedom is urging Senate Commerce to keep the Kids Online Safety Act and the App Store Accountability Act out of a national AI framework, warning that age-verification mandates would funnel children's data into breach-prone databases. Apple previewed a different answer at WWDC — device-level age verification that tells an app a child's age band without sharing a birthday. Same fear, opposite instinct.

Which leaves the question under all of it: the hearing, the bills, the WWDC demo, every syllabus being rewritten this summer. When has policy ever kept pace with the technology it's trying to govern? Usually, we can afford the lag and catch up. Not here: we're being asked to draw the line between the consumer chatbot and the classroom tool before the evidence exists to say where it belongs. This time, the gap is the whole game.

Looking for your next opportunity in education? W/A Jobs features 3,500+ career opportunities from 300+ organizations across the education industry.