AI换脸，移花接木……以后这种视频不能再随意发布了

前一段时间换脸软件“ZAO”一夜爆红，在刷屏朋友圈的同时，也让很多网友对这类换脸技术很是担忧。不过如今，面对这样的AI技术视频，终于有了相关办法出台。

近日，国家互联网信息办公室、文化和旅游部、国家广播电视总局联合印发了《网络音视频信息服务管理规定》（以下简称《规定》）。《规定》中显示，将禁止随意发布AI（人工智能）造假视频，以应对Deepfake换脸等高逼真度图片、视频引发的问题。

The Cyberspace Administration of China (CAC) announced a ban on fake video news produced with artificial intelligence (AI) in a move to address growing concerns over Deepfake technology's dissemination of altered pictures and videos that look convincing enough to be original.



新规定将从2020年1月1日起施行。《规定》中表明：“利用基于深度学习、虚拟现实等新技术新应用制作、发布、传播非真实音视频信息的，应当以显著方式予以标识，不得利用其制作、发布、传播虚假新闻信息；网络音视频信息服务提供者应当部署应用违法违规音视频以及非真实音视频鉴别技术，应当建立健全辟谣机制。”

The ban will take effect on January 1, according to the CAC website. Net users or video platforms will be prohibited from creating, distributing, or broadcasting news content altered by new technologies like deep learning and virtual reality.Net users and platforms are required to noticeably mark AI-altered videos, according to the new CAC rule. Platforms will be asked to implement methods to detect AI-altered videos and timely stop the circulation of unmarked ones. A mechanism should be built to control rumors in case "fake videos" cause "fake news," said the CAC.

此前，网友将94版《射雕》里黄蓉的扮演者从朱茵的脸换成了杨幂的脸，服化道细节逼真，表情也是惟妙惟肖：




此外，《规定》中还显示：“网络音视频信息服务提供者基于深度学习、虚拟现实等新技术新应用上线具有媒体属性或者社会动员功能的音视频信息服务，或者调整增设相关功能的，应当按照国家有关规定开展安全评估。”

The ban will require platforms to conduct risk evaluations before the launch of AI-based video services.

“网络音视频信息服务提供者和网络音视频信息服务使用者违反本规定的，由网信、文化和旅游、广播电视等部门依照《中华人民共和国网络安全法》《互联网信息服务管理办法》《互联网新闻信息服务管理规定》《互联网文化管理暂行规定》《互联网视听节目服务管理规定》等相关法律法规规定处理；构成违反治安管理行为的，依法给予治安管理处罚；构成犯罪的，依法追究刑事责任。”

Violators of the ban will be punished in line with existing laws regulating cyberspace security and cyberspace information services, and may face administrative penalties or criminal prosecution depending on how serious the violation is.



分析人士赞赏该《规则》是一项十分及时的法律行动，说明网信系统跟上了社会潮流以及网络科技的发展。

Analysts hailed the move as a timely legal action, demonstrating the system is keeping pace with social trends and the development of cyber technology.

中国网络空间战略研究所所长秦安表示，作为中国出台的第一部针对媒体中AI技术使用的规则，此举将解决AI换脸技术带来的问题和挑战。秦安表示，新规则具有实践性，并且为视频检测领域开辟了新的市场。

As the first regulation in China on the use of AI in the media, it will address the problems brought about by AI face-changing and other related challenges, said Qin An, head of the Beijing-based Institute of China Cyberspace Strategy, noting that the new regulation is practical and could open up a market for video detection.

此前，网友制作的“黑寡妇&佟湘玉”、“雷神&徐锦江”等视频都曾吸引众多关注。




《环球时报》英文版网站报道指出，新规定的出台，正值人们越发关心换脸技术带来的隐私风险及法律冲突之时。此前在西方国家出现的“换脸明星色情视频”等现象在中国也已出现。

The ban comes in the wake of increasing privacy concerns and legal conflicts regarding Deepfake technology. Deepfake pornography with celebrities' faces, a phenomenon in some Western countries, has emerged in China.

报道指出，现已发现地下卖家通过在线论坛用明星脸部图片吸引买家，贩卖色情制品，并在完成微信支付或支付宝交易后通过云端分享不法链接。

The Global Times previously found that underground sellers of pornography with actors' faces attract buyers at online forums and share videos via cloud drive link after receiving money through WeChat or Alipay.

《华盛顿邮报》2018年曾报道：国外主流社交平台Reddit已禁止使用换脸技术，因为此前诸如泰勒斯威夫特和盖尔加朵等明星的面部图像被用在了色情视频中。



今年9月，换脸软件ZAO爆红后，随即因用户隐私问题道歉，并两次修改了相关条款。此前，ZAO在用户协议中告知，授权app后，ZAO即拥有权力无限期商用用户图像，并且无需用户同意。

In September, the face-changing app ZAO apologized and revised its user terms twice after public accusations of potential abuse of users' privacy. A ZAO terms of use notice told users that by granting permission to the app to access data, they authorized the company's right to use their portraits for commercial use indefinitely and without users' consent.



综合来源：环球时报英文版网站，人民日报官方微博，网信办官网