Open Problem: Protective-Design Pre-Emption in Child-Facing Generative AI

Generative AI is adopted rapidly by children, yet unrestricted general-purpose models dominate their usage. This paper presents the open problem of $\textit{protective-design pre-emption}$: Delayed market introduction, low access thresholds, utility gaps, habituation, and peer adoption foster unrestricted AI model lock-in, leading children to bypass child-centric platforms. With first signs of a lower adoption of child-centric platforms, we argue that this shift may reinforce existing AI risks such as exposing children to deepened parasocial dependencies, skill erosion, and misinformation. To address this problem, we propose interdisciplinary research directions to quantify pre-emption, understand its socio-technical drivers, and develop interventions ensuring child-centric AI is actively adopted, not just available.