Workflow
AI内容标识管理
icon
Search documents
整治AI滥用行动开展逾月,抖音微博小红书等交“成绩单”
Nan Fang Du Shi Bao· 2025-06-12 09:41
Group 1 - The core viewpoint of the articles is the ongoing regulatory efforts in China to address the misuse of AI technology, focusing on cleaning up illegal AI products and enhancing content identification [1][2][3] - The "Clear and Rectify AI Technology Misuse" campaign is divided into two phases, with the first phase emphasizing source governance and the second phase targeting the dissemination of false information and inappropriate content [1][2] - Various platforms, including Weibo and Douyin, have reported significant actions taken, such as Weibo cleaning 162 pieces of content related to AI face-swapping tutorials and closing 22 accounts, while Douyin removed 24,749 pieces of false and vulgar AI-generated content [1][2] Group 2 - The Shanghai Municipal Cyberspace Administration has guided 15 key platforms, including Xiaohongshu and Bilibili, to clean up illegal AI products and related marketing information, resulting in the interception of over 820,000 pieces of illegal information and the disposal of over 1,400 accounts [2] - In the area of training data governance, Baidu has cleaned its data using authoritative sources, while Baichuan Intelligence has stopped using questionable data sources and established strict web scraping regulations [2] - The management of AI content identification is a priority, with new regulations set to take effect on September 1, which will require platforms to implement explicit and implicit identification for AI-generated content [3][5] Group 3 - Nearly 60 companies in Beijing have already implemented content identification requirements, with platforms like Weibo and Douyin achieving dual labeling of generated content [5] - Companies such as MiniMax and Xiaohongshu have completed the implementation of explicit identification standards and are working on implicit identification and verification [5] - MiniMax has developed technical solutions to add metadata for implicit identification in downloadable files and plans to enhance content recognition through additional markers in AI-generated text and audio [5]