加州与纽约在第二轮聊天机器人监管中走上不同路径

New York Governor Kathy Hochul signed S 9051B, prohibiting the provision to minors of twelve categories of "unsafe AI companion features," covering harmful content, claiming to be human or to have emotions, sycophancy, proactive emotional questioning, and personalization using information from more than 12 hours earlier or from past conversations; the law takes effect on January 1, 2027, with a maximum fine of 25,000 US dollars per violation, and an algorithmic disgorgement mechanism may apply that can require destruction of unlawful data and algorithms trained on it. More than a month earlier, California Governor Gavin Newsom signed SB 1119 (Adam's Law), which takes a governance approach, requiring a pre-launch child safety risk assessment, default parental controls (persistent memory and push notifications off by default, a 1-hour limit per session and a 2-hour daily limit), crisis response protocols and an independent audit every two years, with most requirements taking effect on July 1, 2027.