The rapid, nonlinear evolution and cross-domain penetration of artificial intelligence (AI) have precipitated significant governance deficits, manifesting as “timing dilemmas” for regulatory intervention and “miscalibrated intensity risks”. These challenges expose the limitations of traditional static and linear regulatory paradigms, trapping them within the “control dilemma” articulated by the Collingridge dilemma. This paper constructs a dynamic governance framework predicated on the coupling of temporality and proportionality. Through a historical analysis of disruptive technology regulation and a comparative study of the European Union, the United States, the United Kingdom, and China, the research identifies persistent key constraints: insufficient information and limited control, rendering purely precautionary or laissez-faire approaches inadequate. Consequently, the paper proposes a three-stage rolling intervention mechanism triggered by social penetration rate and contextual sensitivity, alongside a four-tier risk-calibrated toolkit grounded in the principle of proportionality. This framework is supported by complementary measures such as evaluation and auditing systems, clear accountability chains, and international procedural mutual recognition, aiming to achieve a dynamic equilibrium between fostering innovation and ensuring safety.
Wei et al. (Sun,) studied this question.