Abstract
The rapid evolution of intelligent media has driven machine learning technologies to deeply integrate into content production. While lowering the barriers to content creation and enhancing dissemination efficiency, these technologies also provide technical support for the generation and spread of false content, giving rise to a series of challenges in social governance, information security, and ethical considerations. Set against the backdrop of the intelligent media era, this paper examines the core technical pathways and manifestations through which machine learning enables the creation of false content, analyzes the diverse risk profiles associated with such content generation, and focuses on addressing the technical bottlenecks and practical challenges in current risk tracing efforts. Ultimately, it establishes a tripartite risk tracing framework encompassing "technical tracing, subject tracing, and regulatory tracing," while proposing targeted optimization strategies to provide theoretical insights and practical guidance for regulating machine learning applications, curbing the spread of false content, and maintaining order within the intelligent media ecosystem.