China’s Practice and Institutional Improvement of Privacy Protection for Embodied AI: From Data Compliance to Contextual Compliance

Abstract

Embodied artificial intelligence introduces distinctive privacy risks through data collection modalities such as continuous sensing and imperceptible capture in physical spaces. The static data-compliance framework anchored in notice and consent proves inadequate in addressing these risks. Drawing on the theory of contextual integrity, this article examines China's legislative, judicial, and regulatory practices in privacy protection for embodied AI, and identifies two systemic deficiencies: the lack of context-specific rules and weak contextual adaptability. It then proposes a governance paradigm shift from data compliance to contextual compliance, which entails context-specific legislation, a full-lifecycle regulatory architecture spanning ex ante, intra-process, and ex post stages, refined technical standards for context-based risk classification, and strengthened international cooperation, thereby establishing a differentiated and dynamic privacy protection mechanism. The study aims to provide an institutional framework that reconciles technological innovation with the protection of personality rights, while offering insights from Chinese practice for the global governance of embodied AI.