{"p"=>"The diffusion of deepfake technology has reshaped the structure of online risks and posed new institutional demands on existing network security management obligations. From a risk-oriented perspective, this article constructs a quantitative model for assessing artificial intelligence misuse risks and employs simulation-based experiments to systematically analyse governance outcomes under different levels of management obligations. The findings indicate that response timeliness, institutionalized procedures, and evidence preservation play a decisive role in damage mitigation, while reliance on technical detection alone is insufficient to achieve effective governance. The simulation model assesses the four types of governance systems for deepfake misuse across different scenarios through 10,000 iterations of the simulation experiment. Based on the experimental results, the article proposes a risk-tiered configuration of network security management obligations, providing empirical support for judicial determinations of the duty of reasonable care and for policy-level collaborative governance."}