详解标识新规：如何给AI内容强制上“户口”

Core Viewpoint - The implementation of regulations and standards related to AI-generated content aims to address the increasing issue of misinformation and content pollution on the internet, providing a framework for identifying AI-generated content and ensuring accountability among creators and users [1][5]. Group 1: Explicit Identification - All AI model or application providers must add explicit identification when generating content, clearly indicating that it is AI-generated [2][3]. - For text content, explicit identification must be placed at the beginning, middle, or end, using terms like "AI" or "generated" [2]. - Image content requires identification in the corner, with text height not less than 5% of the shortest side of the image [2]. - Audio content must announce "generated by AI" at the beginning or end, or use a specific Morse code rhythm [2]. - Video content follows similar rules to images, requiring identification in the corner and lasting at least 2 seconds [3]. - All interactive interfaces of AI models or applications must also display identification [3]. - Organizations and individuals are prohibited from maliciously deleting or altering these identifications [3]. Group 2: Implicit Identification - Implicit identification serves regulatory purposes and must be included in the metadata of generated content, containing information such as whether it is AI-generated, who generated it, and a unique content number [6][7]. - The regulations encourage the addition of digital watermarks to enhance traceability [6]. - The requirement for implicit identification is mandatory, while digital watermarks are encouraged but not compulsory [7]. Group 3: Compliance and Responsibilities - Product entrepreneurs and AI tool providers are responsible for ensuring compliance with the new regulations, including automatic addition of explicit and implicit identifications [8][9]. - User agreements must clearly inform users about the existence and legal requirements of these identifications [9]. - If services are provided through APIs, the service provider must ensure that implicit identification is included in the generated data [10]. - Users must declare and label any AI-generated content, regardless of the proportion of AI involvement [10][11]. - Actions to remove implicit identification may lead to penalties, including reduced account credibility [11]. Group 4: Impact on Content Creation - The regulations are favorable for serious content creators while posing challenges for those spreading misinformation [11]. - The introduction of implicit identification, such as digital watermarks, serves as a deterrent against the spread of false information and enhances content accountability [11].