<p>Deepfake, which uses artificial intelligence to create highly realistic but false content, can spread rapidly through the Internet, causing damage to both individual rights and social interests. In response, the United States, the European Union, and China have taken the application scenario, life cycle, and responsible subjects of deepfake as the regulatory focus point, thus forming three different regulatory paradigms. The United States adopts an application-based regulatory paradigm, primarily focusing on the use of deepfakes in elections and pornography. The European Union regulates deepfake at different phases through multiple laws and regulations, presenting a life cycle-based regulatory paradigm. China takes the service provider of deepfakes as the subject of regulation and requires it to bear a series of responsibilities, forming a subject-based regulatory paradigm. While these three paradigms deter and mitigate illegal deepfake activities to some extent, they have inherent limitations. From the perspective of the digital ecosystem, current regulation only focuses on content producers, service providers, and distribution platforms but overlooks the significant role of the deepfake audience. By improving the information literacy of the audience, it is possible to enable the audience to effectively identify deepfakes, curtail the spread of harmful and illegal deepfakes, and minimize their impact on victims, thus striking a balance between reducing risk and promoting technological development.</p>