9月12日下午，以"规范AI内容 共筑清朗生态"为主题的2025外滩大会见解论坛上，蚂蚁数科AI科技产品 总监鲁玮发表主题演讲，从身份伪造、凭证造假等AI滥用风险问题切入，系统介绍蚂蚁在应对生成式 内容风险、黑灰产攻击以及大模型"幻觉"方面的技术实践与治理思考。 展望未来，鲁玮认为，随着具身智能、AI for Science与智能体（Agent）加速落地，AI能力跃升将伴随 更复杂的安全议题，包括人机互动的责任边界、价值对齐与伦理问题等。他呼吁产业、学界与媒体加强 协同，共建规范、透明、可信的AI治理体系。 本次论坛由南方都市报社、南都大数据研究院、中国互联网协会人工智能工作委员会、复旦大学传播与 国家治理研究中心联合主办。 目前，黑产借助AI的凭证造假愈发低成本、规模化。在凭证防伪方面，鲁玮介绍，蚂蚁数科推出"光鉴 智能验真"，通过防伪、质检、识别，最后通过高阶的各种决策，大小模型结合，实现在语义、视觉、 文本、图片多重信息的抽检，最后给出关键的决策，"核心逻辑是保证现在的物理世界里的票据、身 份、凭证是防伪的。" 幻觉同样是大模型应用面临的重要挑战之一。鲁玮认为，AI生成内容幻觉的危害在于滥用人类的信 任 ...

中新网上海9月12日电 (姜煜高志苗)以"AI向善·安全发展——大模型在严谨行业应用中的伦理与治理"为 主题的外滩大会分论坛，12日在上海"2025外滩大会"现场举办，专家学者在论坛上热议医疗AI应用。 "在医疗行业，AI的决策直接关乎生命健康与社会公平，数据滥用、算法偏见、责任模糊等问题亟待解 决。"中国互联网协会副理事长黄澄清在论坛致辞中表示，构建可信、可靠、可追溯的AI治理体系已成 为技术发展的内在要求，也是社会责任的必然选择。 "医疗大模型的发展，我觉得对社会最大的贡献是有可能促进医疗公平公正。"复旦大学哲学学院教授王 国豫说，AI在医疗领域的应用是偏远地区医生的好助手，为优势医疗资源难以触达和覆盖的病患群体 带去了期望和福音。 上海交通大学医学院附属瑞金医院信息中心主任赵艳认为，AI的出现让患者不仅在诊前就能对自身情 况进行较科学的基础了解，也在诊中和诊后促进了患者对医生诊疗方案的依从性，从而可以提升配合治 疗的效果。 企业界代表、蚂蚁集团资深副总裁及科技伦理委员会联席主席周志峰说："我们还是希望通过预警、测 试、修复这样的动态机制来提升AI的风险防控能力。"他在论坛上介绍了蚂蚁集团在AI大模型领域 ...

Core Insights - The widespread application of generative AI technology and explosive growth of digital content have reshaped the boundaries of intellectual property protection, calling for more innovative and systematic governance paradigms [1] - Taotian Group has established a new paradigm for intellectual property protection driven by "innovative protection and infringement governance," laying a solid foundation for the high-quality development of the digital economy [1] Group 1: Intellectual Property Protection Initiatives - Taotian Group's "proactive prevention" project has quickly identified and halted infringements, achieving rapid online governance effects [1] - The company has launched five exclusive rights for small and medium-sized enterprises (SMEs) and emerging brands, including "consultation services," "rapid rights protection," "brand co-construction," "offline cooperation," and "original protection" [2][3] - The platform has supported over 240 SMEs and emerging brands, with the volume of governance on infringing products increasing by 49% [3] Group 2: Original Protection Platform - The Original Protection Platform, launched in 2018, has protected over 897 million images, 151 million short videos, and more than 930,000 design manuscripts, benefiting over 960,000 original merchants by March 31, 2025 [4] Group 3: AI Governance Mechanism - Taotian Group has established an "innovative protection" and "poor quality governance" dual-driven AI governance system, implementing a first-release filing mechanism for AI-generated content [5] - The platform's AI image protection mechanism allows merchants to generate a "digital ID" for each AI design image, enabling quick rights protection through filing evidence [6] Group 4: Global Cooperation and Brand Protection - Taotian Group has created a "global identification cooperation alliance," with over 1,000 brands expressing interest in collaboration to enhance the identification of infringing products [7][8] - The alliance aims to build an intelligent system for identification cooperation, reducing the cost of brand self-enforcement and focusing on high-risk categories and products [8]