This is a quick follow-up on my post about NYC's new AI policy.

I've been digging into how NYC's policy compares to LA Unified's similar moratorium, and one difference stands out. LAUSD's ban targets student devices, teachers keep their AI access. NYC went further and put teachers on the red list too.

One practical difference? NYC allows no AI for grading; personalized feedback got swept up right alongside high-stakes calls like promotion and graduation.

That's a tragic mistake. Student feedback and promotion decisions are not the same category of risk. Nobody's diploma is on the line because a model helped draft a comment on an essay, but every teacher I've ever talked to would kill for the time to give that kind of feedback to all 150 students instead of just the dozen who caught a spare period.

The real driver here isn't feedback specifically, it's that "any decision about a student" got defined broadly enough to swallow it. Blanket caution is easier to defend than a policy that actually distinguishes "helped a kid with fractions" from "decided if they graduate."

In a Wall Street Journal opinion piece this past weekend, Ravi Gupta writes about AI resistance in schools and wryly notes that "AI tutors don't pay union dues". And he got me thinking about how the teachers unions may have shaped this (perhaps without quite meaning to). LAUSD's contract with UTLA bars AI from replacing bargaining-unit work, but that's a job-protection clause, not an assessment one, so a teacher using AI to help with feedback doesn't trip it. NYSUT pushed for something similar in spirit, AI shouldn't replace a teacher's judgment, but NYC seems to have run with that principle much further, to where AI can't touch a decision at all, human in the loop or not. Same underlying worry, two very different policies.

As I noted in the previous post, I do understand the need for caution, but, by preventing teachers from using a tool that will help them help kids, NYC has gone too far. A ban that can't tell fractions from graduation isn't protecting kids, it's protecting itself.