AI反欺诈
Search documents
浩瀚深度20251128
2025-12-01 00:49
浩瀚深度 20251128 摘要 浩瀚深度收购国睿数智 35%股权并实现并表,后者 2024 年净利润 900 多万元,目标 2025 年达到 3,500 万元,尽管业绩波动大,但公司对其 全年业绩充满信心,并已通过可转债再融资为国睿提供资金支持。 国睿数智与浩瀚深度在客户群体上形成互补,前者服务于国家网络监管 部门,后者服务于运营商,双方技术互补,例如在反诈领域,浩瀚提供 前端采集设备,国睿负责后端软件分析系统。 浩瀚深度对 AI 业务前景乐观,预计 2026 年将是重要发展年份,正与工 信部、信通院及三大运营商合作推进多个重点项目,并计划与国睿数智 合作开发基于 AIGC 的新型监管解决方案。 反诈市场潜力巨大,浩瀚深度与国睿数智将整合资源,推出完整解决方 案,已在辽宁管局项目中获得成功,计划拓展至其他地区及行业客户, 并持续进行技术创新和产品升级。 语音反诈和互联网反诈的主要区别在于采集和识别方式,国瑞在语音反 诈方面经验丰富,浩瀚在互联网反诈方面拥有 DPI 技术优势,结合大数 据模型提供全方位解决方案。 Q&A 浩瀚深度为何决定进一步收购国睿数智的控股权? 浩瀚深度决定进一步收购国睿数智的控股权,主 ...
Soul App发布第三季度生态安全成果
Zheng Quan Ri Bao Wang· 2025-10-23 11:15
Core Insights - Soul App has released its "2025 Q3 Ecological Safety Report," highlighting advancements in ecological safety and showcasing its innovative practices in user information protection, anti-fraud measures, minor protection, online violence governance, and community co-governance [1][2]. Group 1: AI Technology and Safety Measures - The platform has optimized its AI-based image anti-fraud model, achieving an approximately 80% reduction in false positives and increasing recognition coverage to over 90%, ensuring comprehensive protection against mainstream image production scenarios [2]. - The high-risk user identification coverage has improved from below 50% to over 70%, enhancing the platform's ability to preemptively warn and identify potential fraudulent users [2]. - Soul App has maintained close cooperation with regulatory authorities and law enforcement, contributing over 100 reports of illegal harmful information and participating in multiple collaborative efforts to combat such content [2]. Group 2: Content Moderation and Community Engagement - In Q3, the platform intercepted 4.39 million pieces of violating content through automated systems, processed 15.78 million related comments, and blacklisted 30,600 violating accounts, effectively curbing the proliferation of inappropriate content [2]. - The community governance initiative has seen over 78,000 users participating as content recommenders and discipline committee members, with daily contributions of over 22,000 votes leading to the generation of more than 2,900 quality posts [3]. - The report serves as a commitment to users and society, demonstrating Soul's dedication to building a robust safety defense through AI technology, strict enforcement, and community collaboration [3].