10月25日，第32届中国国际广告节举办期间，巨量引擎商业安全开放日作为分会场同步召开，北京市市场监管局、中国广告协会相关领导出席并讲话。会 上，巨量引擎介绍其 "全链路治理+以AI治AI" 的广告安全防线构建核心策略，公布了今年第三季度已前置拦截84万余个涉AI广告违规素材。首次披露AI广 告治理领域自研多模态大模型，该模型可实现90%的素材10分钟内完成审核，为AI广告风险治理提供技术支撑。 潘荻在现场分享道，与传统依赖案例训练的 "判例法" 模型不同，加入AI能力之后的大模型，采用的是 "条文法" 识别逻辑，能结合大量案例学习，学会问题 推理过程，从而更好地理解规则。这种能力升级直接带来了审核精度的显著提升，一方面大幅降低了因案例覆盖不全导致的误判、漏判风险，另一方面让审 核结果更贴合规则本质。 值得关注的是，广告后链路履约环节虽已脱离平台管控，但其服务质量直接影响用户对广告生态的信任，尤其是老年人、未成年人等群体，在维权操作、风 险识别上更易遇到障碍。为解决这一痛点，巨量引擎专门推出针对一老一少的专项服务保障项目。据郭泽斌介绍，当用户因抖音广告引流后引发了消费纠 纷，平台可提供先行保障等服务。用户可直 ...

构建全链路管控84万+条违规素材被拦截 面对AI广告的三大突出风险——违背公序良俗、侵权抄袭、虚假宣传，巨量引擎已建立全流程治理体系。在活动现场，巨量引擎商业安全治理标准负责 人项韵介绍，平台聚焦"精准识别、从严处置"，例如要求广告内容中不得含有虚构人设，不得自称、或描述存在某某身份/姓名/职位，为产品或服务做背 书。同时，针对违规行为设置了从预警整改、限制新开、账户封停、主体清退等梯度化、多元化处置体系。据悉，今年第三季度已前置拦截涉AI违规素 材84万余个。 10月25日，第32届中国国际广告节举办期间，巨量引擎商业安全开放日作为分会场同步召开，北京市市场监管局、中国广告协会相关领导出席并讲话。会 上，巨量引擎介绍其"全链路治理+以AI治AI"的广告安全防线构建核心策略，公布了今年第三季度已前置拦截84万余个涉AI广告违规素材。首次披露AI广 告治理领域自研多模态大模型，该模型可实现90%的素材10分钟内完成审核，为AI广告风险治理提供技术支撑。 自研多模态大模型审核效率提升75% "技术是治理的核心驱动力。"巨量引擎商业安全产品负责人潘荻展示的交互式审核系统，依托自研多模态大模型实现三大突破：90%素材 ...

Core Insights - The core focus of the news is on the advertising safety measures implemented by the company, particularly in the context of AI-generated advertisements, highlighting their proactive strategies and technological advancements in governance [2][3][4]. Group 1: Advertising Safety Measures - The company introduced a comprehensive governance system to address three major risks associated with AI advertising: violation of public morals, copyright infringement, and false advertising [2]. - In the third quarter of this year, the company successfully preemptively intercepted over 840,000 AI-related advertising violations [2]. - The company has established a multi-tiered response system for violations, including warnings, account restrictions, and permanent bans [2]. Group 2: Technological Innovations - The company unveiled a self-developed multimodal model capable of reviewing 90% of advertising materials within 10 minutes, improving review efficiency by 75% compared to traditional methods [3][4]. - A new Consumer Complaint Rate (CCR) index was introduced, which integrates user feedback and has seen a reduction of over 40% in negative feedback regarding advertisements this year [3]. - The interactive review system can automatically correct common violations in advertising materials and provide detailed rejection reasons, enhancing the efficiency and experience for advertisers [3][4]. Group 3: Targeted Support Initiatives - The company launched specialized service projects aimed at protecting vulnerable groups, such as the elderly and minors, by providing preemptive support in case of advertising-related disputes [5]. - The platform has conducted offline anti-fraud lectures for the elderly and created educational content for minors to raise awareness about advertising risks [5]. - The specialized service has assisted over 3,000 individuals, achieving a user satisfaction rate of 92.5% [5].