Soul IPO进程持续，生态安全报告凸显平台治理成果

Core Viewpoint - Soul's recent ecological safety report for 2025 highlights its advancements in fraud prevention and content safety, aiming to assure investors and users of its systematic capability to maintain a healthy ecosystem during its IPO process [1][9] Group 1: AI Integration in Governance - In 2025, Soul plans to deeply integrate AI into its governance cycle, deploying seven self-developed AI models to create a comprehensive safety shield [3] - The "Image/Text Anti-Fraud Model" can understand deep semantics and potential intentions, identifying not only surface-level violations but also the underlying logic of fraudulent activities [3][6] - Soul has developed a unique "Pig Butchering Posting Recognition Model" to target emotional manipulation and persona deception, allowing for early interception of fraud attempts [3][5] Group 2: Combatting False Traffic - Soul upgraded its governance matrix to address issues of "fake personas" and "image theft," achieving an average daily interception of 50,000 false content pieces through AI image path analysis and device anomaly detection [3][7] - This "circuit breaker" mechanism effectively disrupts the breeding ground for fraudulent activities, ensuring that users encounter content that reflects real life [3][7] Group 3: Collaborative Efforts with Law Enforcement - In 2025, Soul deepened its collaboration with law enforcement agencies, including the Pudong Anti-Fraud Center, by submitting over 100 illegal activity leads and assisting in investigations more than 30 times [4][8] - The company is building a closed-loop system of "in-platform interception + external crackdown" to create a safety network that leaves no room for wrongdoers [4] Group 4: Impact on User Ecosystem - The introduction of the "AI Anti-Fraud Plan" includes six recognition models and an AI reminder system to enhance user awareness and create a dual defense against fraud [5] - The number of individuals involved in fraudulent activities has decreased by approximately 20% compared to 2024, showcasing the effectiveness of the AI models in combating various forms of fraud [6]