Beyond Parental Control: A Human-Centred AI Framework for Child-Safe Online Video and Gaming

Child online safety is commonly addressed through parental supervision, age restrictions, and content filtering. This paper argues that safety should also be treated as a platform-level property of online video and gaming systems. We introduce the Child-Safe AI Decision Architecture (CS-AIDA), a human-centred framework that connects age baselines, optional safety-competency signals, multi-dimensional contextual risk assessment, graduated intervention, explanations, escalation, and continuous audit across content, interaction, recommendation, behavioural, and privacy risks. As a secondary contribution, we propose the Digital Maturity Passport, a researchable mechanism for linking validated, demonstrated safety competencies to bounded increases in digital autonomy while preserving non-negotiable protections against severe harm. We distinguish architectural integration from the invention of individual technical components and position CS-AIDA as an evaluable design framework rather than a validated production system. The paper contributes a structured harm taxonomy and a five-stage empirical programme covering architecture validation, competency-instrument development, controlled intervention comparison, longitudinal field evaluation, and fairness and robustness auditing. The guiding principle is: protect first, teach continuously, and increase autonomy as demonstrated competence develops. The framework is intended to support research in HCI, responsible AI, recommender systems, and child online safety.