（记者 王瀚黎） 每经头条（nbdtoutiao）——AI技术滥用调查：明星可被"一键换装"，"擦边"内容成流量密码，技术防 线为何形同虚设？ 每经AI快讯，永清环保10月13日发布公告称，永清环保股份有限公司董事会于近日收到公司证券事务 代表黄田女士的书面辞职报告，黄田女士因个人原因申请辞去公司证券事务代表职务，辞职后将不再担 任公司及子公司任何职务。 ...

一段"AI换脸"的不雅视频，让一位高校导师突然陷入诈骗危机；一张"近乎复刻"的AI克隆照，让白领小雅担心自己的照片被用来制作色情内容；青岛保时 捷女销冠和台球运动员王思诺则深陷恶意AI造"黄谣"风波……他们都是AI技术被滥用的受害者。 此外，社交平台上，明星"AI换装""AI擦边"内容泛滥，成为"流量密码"，部分账号靠此类内容吸粉，甚至有教程指导用户借此"起号变现"。 这些涉嫌侵权和违规内容缘何能被轻易炮制出来？ 近期，《每日经济新闻》记者（以下简称"每经记者"或"记者"）对用户使用较多的12款文生图和文生视频AI应用进行了实测，结果发现5款可实现明星"一 键换装"，9款能生成"擦边"图片。 面对乱象，中央网信办于今年6月启动了"清朗·整治AI技术滥用"专项行动第二阶段工作，重点整治7类突出问题，包括利用AI制作发布色情低俗内容，利 用AI假冒他人实施侵权违法行为。 然而，AI应用的技术防线为何脆弱不堪？内容平台又在其中扮演了何种角色？这场技术滥用乱象背后的多重疑问，亟待解答。 一、AI滥用阴影下：普通人与公众人物皆难幸免 8月的一个深夜，电子科技大学人工智能专业研究生校外导师高翔收到一条诈骗短信，短信中是一 ...

Group 1 - The article highlights the misuse of AI technology, particularly in creating inappropriate content and identity theft, affecting both ordinary individuals and public figures [2][4][6] - A recent investigation tested 12 popular AI applications, revealing that 5 could easily perform "one-click dressing" of celebrities, while 9 could generate suggestive images [26][27][31] - The prevalence of AI-generated content on social media platforms has led to a surge in accounts exploiting this technology for gaining followers and monetization [7][8][21] Group 2 - The article discusses the weak defenses against AI misuse, questioning the role of content platforms in preventing such abuses [3][36] - Legal frameworks exist to regulate AI-generated content, but there are challenges in enforcement and clarity regarding "borderline" content [39][40] - Experts suggest that improving detection technologies and increasing penalties for violations could help mitigate the misuse of AI [38][41]

央视网消息（焦点访谈）：当下，在短视频平台或者是直播平台上购物成了很多人的新选择。刷视频的时候，还经常会冒出熟悉的明星，用流利的乡音推销 农产品，有的还进行直播带货。注意了，这些所谓的"明星"很可能是仿冒的。近段时间，网络上一些商家利用AI技术仿冒带货。有的借助名人效应，仿冒专 家学者、奥运冠军、影视明星；也有的借助产地效应，冒充知名品牌。消费者、粉丝难辨真伪，慕名下单，却不知道自己被蒙骗了。这些游走在灰色地带 的"AI李鬼"，正拷问着我们的数字治理能力。 近段时间，在一些手机平台的网络直播间，一些打着樟树港辣椒旗号的种子，成了主播推销带货的重点。记者注意到，这些直播间的销售人气很旺，不时有 人咨询或下单购买。 湖南岳阳的樟树港辣椒属国家地理标志农产品，有400多年的种植历史，因量少质优，特级产品可卖到300元一斤。 而在这些网络直播间，记者看到，画面中展示的辣椒成串垂挂，产量奇高。当记者在直播间询问辣椒种子是否保真，种出来有没有这么大时？主播纷纷表 示，所见即所得。 奇怪的是，就在主播回答提问的时候，背景视频却出现了农户站在快要没过膝盖的深水中采摘辣椒的场景，这样的场景显然不符合常理，而类似的画面在其 他直 ...

Core Viewpoint - The proliferation of AI technology has enabled the "cloning" of specific individuals' voices, leading to civil infringement and potential criminal issues, particularly in the context of online marketing and live streaming [1][5]. Group 1: AI Voice Cloning in Marketing - Some social media influencers are using AI-cloned voices of Olympic champions, such as Quan Hongchan, to promote agricultural products, misleading fans into believing they are interacting with the actual athletes [3][5]. - A specific account on Douyin has posted 17 videos using AI to impersonate Quan Hongchan, achieving over 11,000 likes on one video, and selling 47,000 units of eggs linked to the promotion [3][5]. - Other Olympic champions, including Sun Yingsha and Wang Chuqin, have also been impersonated in similar marketing schemes [5]. Group 2: Legal and Ethical Concerns - The use of AI to clone voices raises significant legal concerns, including potential fraud and infringement of personal rights, as highlighted by experts [6][7]. - China's Civil Code now includes provisions for the protection of an individual's voice, equating it to portrait rights, thus making unauthorized use of a person's voice for AI cloning an infringement [6][7]. - Experts recommend that platforms implement stricter regulations and monitoring mechanisms to prevent the misuse of AI voice cloning technology [7]. Group 3: Regulatory Measures - In March, China's National Internet Information Office and other departments announced a regulation set to take effect in September 2025, requiring explicit labeling of AI-generated content [7]. - A nationwide campaign has been launched to address the misuse of AI technology, focusing on enhancing detection and verification capabilities on platforms [7].

随着AI技术的普及，"克隆"特定人物声音已实现"一键式"操作，同时也带来了民事侵权甚至涉嫌犯罪的 法律问题。 在互联网上，奥运冠军的"声音"突然开始推销农产品，知名演员的"声音"24小时不间断在直播间与粉丝 互动……这些以假乱真的AI克隆声音正成为部分自媒体博主的"流量密码"。 记者在抖音平台发现，在这个网名为"我是小助理吖"的博主账号中，已发布的17件作品，内容全部是利 用AI仿冒奥运冠军全红婵声音带货土鸡蛋的视频，其中一个视频点赞量高达1.1万。 评论区显示，大量粉丝误以为这就是全红婵本人，纷纷下单购买链接中的鸡蛋，该商品链接显示已售出 4.7万件。今年4月，全红婵家人曾向媒体证实，全红婵并未给相关农产品代言或授权。 除全红婵外，奥运冠军孙颖莎和王楚钦也遭遇了同样的仿冒。在这个网名为"小兔真可爱"的账号中，两 位奥运冠军的AI克隆声音同样在为土鸡蛋带货。 据悉，依托成熟的语音大模型技术，仅需十几秒就能"克隆"任何人的声音，当声音能被轻易窃取和滥 用，随之而来的侵权乱象频繁上演。 以假乱真 用奥运冠军的AI克隆声音带货 网传AI仿冒全红婵带货画面：大家好，我是你们的婵宝，今天有事请大家帮忙。婵妹就想 着帮助 ...

Core Viewpoint - The rise of AI technology has enabled the cloning of voices, leading to frequent infringement issues and the emergence of a gray industry behind it [1][2][10]. Group 1: AI Voice Cloning in E-commerce - AI has been used to impersonate Olympic champion Quan Hongchan's voice for selling farm products, with one video achieving 11,000 likes and 47,000 items sold [2][5]. - Other Olympic champions, such as Sun Yingsha and Wang Chuqin, have also been victims of similar voice cloning for selling products [4]. - The deceptive nature of AI-generated voices has led to significant consumer confusion, with many believing they are interacting with the actual celebrities [5]. Group 2: Gray Industry and Legal Implications - The use of AI to clone voices has created a gray industry where individuals can quickly gain followers and monetize their accounts through impersonation [8]. - Voice actors are also facing infringement issues, with cases of their voices being used without consent in promotional materials [10][12]. - Legal experts indicate that unauthorized use of AI-generated voices constitutes an infringement of personality rights, as outlined in China's Civil Code [17]. Group 3: Regulatory and Platform Responsibilities - Experts suggest that platforms must take responsibility for regulating AI voice cloning, establishing mechanisms for reporting and reviewing infringing content [19]. - A new regulation set to take effect in September 2025 will require service providers to label AI-generated content clearly [21]. - The Chinese government is initiating actions to combat the misuse of AI technology, focusing on enhancing detection and verification capabilities on platforms [21].

Core Viewpoint - The widespread use of AI technology for voice cloning has led to significant legal issues, including civil infringement and potential criminal activities, as individuals exploit this technology for personal gain [1][17]. Group 1: AI Voice Cloning and Its Applications - AI voice cloning technology allows for the rapid and realistic imitation of any individual's voice, requiring only a short audio sample to produce a convincing clone [13][14]. - Some social media influencers are using AI-cloned voices of famous athletes, such as Olympic champions, to promote products, misleading fans into believing they are interacting with the actual individuals [2][4][8]. - Instances of AI voice cloning have resulted in significant sales, with one influencer reportedly selling 47,000 units of a product while impersonating an Olympic champion [4]. Group 2: Legal and Ethical Implications - The misuse of AI voice cloning not only deceives consumers but also infringes on the personal rights of the individuals whose voices are cloned, raising serious ethical concerns [9][19]. - Legal experts highlight that the Chinese Civil Code now explicitly protects individuals' voices, equating it to portrait rights, thus making unauthorized use of someone's voice a potential legal violation [17][19]. - The legal framework indicates that any use of a person's voice without consent constitutes infringement, emphasizing the need for clear permissions regarding voice cloning [19]. Group 3: Industry Response and Regulation - Experts suggest that platforms hosting AI voice cloning content should bear responsibility for monitoring and preventing misuse, as they can be held liable if they fail to act against infringing activities [20][22]. - The Chinese government has initiated measures to regulate AI technology misuse, including a new directive requiring clear labeling of AI-generated content, set to take effect in September 2025 [22]. - There is a call for platforms to establish robust mechanisms for reviewing and reporting AI voice cloning incidents to mitigate the spread of fraudulent content [22].

魔高一尺，道高一丈。考虑到如今AI造假能力之强，要彻底铲除AI造假"起号"的灰色产业链，关键 在于以技术优势形成绝对压制。相较于不法分子，监管部门与网络平台手握更雄厚的技术资源、更完整 的数据链条和更强大的研发能力，能够通过技术赋能实现对新技术违法犯罪的"冒头即打、露头就灭"。 平台需构建"识别—拦截—溯源"技术体系，用深度模型精准识别伪装的AI内容，建立违规账号"黑名 单"，定向阻断变现链条，让造假者无利可图。监管部门可搭建全国性监测平台，追踪账号交易异常， 解码隐蔽交易信息，联动打击产业链节点。同时，要畅通技术化举报渠道，形成全民监督合力，让灰色 产业链失去滋生土壤。 AI的价值在于拟人化服务而非拟人化欺骗，在于赋能发展而非助力乱象。我们期待以专项行动为 契机，加快推动AI造假治理向以技术对技术、以技术管技术的进阶，让数字空间真正成为滋养文明的 沃土。 来源：检察日报 治理此类乱象，面临着多重现实挑战。一方面，技术对抗的"猫鼠游戏"持续上演，造假者通过压缩 文件、修改代码等方式规避平台AI识别，增加了监管难度；另一方面，法律适用存在模糊地带，AI生 成内容的标注义务、虚拟账号的归属等问题缺乏细化规定，导致 ...

来源：法治网微信公众号 01 法治网研究院：针对近来不断被媒体曝出的换脸、拟声等各种AI技术滥用乱象，您认为，当前整治 AI技术滥用的难点在哪里？ 黄尹旭：当前，AI滥用手段的进化速度可能超过检测技术，深度伪造已从简单的面部替换升级到动态 表情、光影细节的模拟，有些经过对抗性训练绕过现有检测工具。检测系统刚识别出像素异常特征，滥 用者就可能添加噪声、调整参数来规避。 滥用内容的隐蔽性越来越强，可能只是在真实视频中替换几秒对话，或在新闻里夹杂小段AI生成文 字，传统批量扫描技术很难识别局部篡改。 还有，各平台的监测标准、处置流程不一，难以形成全链条拦截。 李瑞：我认为主要难点有三个：一是技术迭代迅猛，随着人工智能技术的飞速发展，深度伪造技术的质 量不断提升，已经日益逼近甚至超越真实水平，而传统的识别模型在这种高速变化面前显得力不从心， 难以跟上技术发展的步伐。 二是责任主体呈现碎片化状态，从最初的训练数据收集、模型托管服务、分发平台运营，到最终的终端 用户使用，整个链条极为冗长且复杂。这种多环节、多主体的分散格局，使得取证过程变得异常艰难， 归责问题也变得错综复杂，难以厘清。 三是治理标准存在明显缺位，现行的《 ...