New York enacts ban on 'unsafe' chatbot behavior with kids - ABC News

Twelve. Not three, not five — a full dozen definitions of “unsafe AI companion features,” inscribed into New York statute by Gov. Kathy Hochul on Wednesday, so that any chatbot addressing a seventeen-year-old may be measured against a list long enough to double as a pub quiz. One pictures the drafting session: a whiteboard, a thesaurus, and somebody's nephew who has “done AI.” Twelve is, to be fair, more definitions than most industries volunteer in a decade of self-regulation, which tells you precisely how much self-regulation was on offer.

And the reason it happened is printed in ABC's own sentence: “With Congress doing little to regulate AI.” Little is diplomatic; the federal contribution to child-safe chatbots currently amounts to an aspiration and a press release. So the states are improvising with consumer protection law, which is the legislative equivalent of fixing the boiler with duct tape and a sincere belief in tort. Delaware, meanwhile, has been off attempting something far grander with AI and corporate personhood, which is the same absence of grown-ups in a different aisle.

Then note the last item on the list: “rogue AI frontier models,” tucked at the end like a wasp in a picnic basket. Companion chatbots are the headline; the frontier is the subtext, and the definitions are being written by people who will never see a model weight, applied to products that ship updates on a Tuesday, enforced by whichever agency draws the short straw. Still, twelve definitions beats the federal alternative, which is nothing dressed up as thinking. One hopes the twelve have sub-clauses; one fears they were drafted in a fortnight by a working group with a deadline and a lunch order. A dozen definitions of unsafe, and Congress is presumably still workshopping the first.