AI Labeling Regulations: The Starting Point of a "Truth" Defense Campaign
人工智能标识新规：一场“真实”保卫战的起点
The recent rumor that "A popular actor lost his $1 billion fortune gambling in Macau" sparked a widespread Internet public opinion storm , later exposed by cybersecurity authorities as fabricated by netizen Xu Mouqiang using AI tools combined with trending keywords. This incident is just the tip of the iceberg of AI misuse—from AI-generated images of "children in ruins" during the earthquake in Shigatse, Tibet, to synthetic videos showing Hollywood’s landmark engulfed in flames, and even Jiangxi-based Wang Moujiang operating five MCN（Multi-Channel Network）agencies to mass-produce 4,000–7,000 AI-generated rumors daily for profit. The "double-edged sword" of technology is eroding trust in the information age at an unprecedented pace.
近日，一则“某超一线男星在澳门输光10亿身家”的谣言引发互联网舆论风波，最终被公安网安部门查明为网民徐某强利用AI工具输入社会热点词制作的虚假信息。这不过是AI技术被滥用的冰山一角——从西藏日喀则地震中被AI捏造的“废墟儿童”影像，到美国著名地标“好莱坞”（HOLLYWOOD）广告牌矗立在火海中的合成视频，再到江西男子王某江经营5家MCN机构组织团队单日量产高达4000至7000篇的AI谣言牟利，技术的“双刃剑”效应正以空前速度动摇信息社会的信任根基。
Against this backdrop, China’s newly released Regulations on Labeling AI-Generated Synthetic Content (hereafter the Regulations), effective September 1, marks a critical step in AI governance—not only a systemic response to technological abuse but also a pioneering effort to rebuild a trustworthy information ecosystem.Jointly issued by the Cyberspace Administration of China, Ministry of Industry and Information Technology, Ministry of Public Security, and National Radio and Television Administration, the Regulations mandate comprehensive labeling obligations across the AI content lifecycle. Three pillars define its framework:Dual-track identification requiring both visible markers (e.g., "AI-generated" watermarks) and invisible digital fingerprints embedded in metadata;Chain-of-custody accountability assigning responsibilities from creators to redistributors, including platforms' duty to verify labels before dissemination;Anti-tampering safeguards prohibiting unauthorized removal or alteration of identifiers, and violations will be dealt with in accordance with the provisions of relevant laws, administrative regulations, and departmental rules.
在此背景下，《人工智能生成合成内容标识办法》（以下简称《办法》）的出台，标志着我国在人工智能治理领域迈出了关键一步。这项将于9月1日施行的新规，既是对技术滥用问题的制度性回应，更是在为重建真实可信的信息环境探路。这项由国家互联网信息办公室、工业和信息化部、公安部、国家广播电视总局四部门联合发布的《办法》，在AI内容的全生命周期中构筑起三重防线：双轨识别，要求同时使用可见标识（例如“AI生成”水印）和嵌入元数据中的不可见数字指纹；责任链追溯，将责任从创作者延伸至再分发者，包括平台在传播前检验标识的义务；防篡改保护，禁止未经授权擅自移除或更改标识，违规行为将依据相关有关法律、行政法规、部门规章的规定予以处理。
The most groundbreaking innovation of the Regulations lies in establishing a dual "explicit + implicit" labeling system. Explicit labels in the generation of composite content or interactive scene interface to forcibly remind the audience by text, sound, graphics and other ways, while implicit labels embed traceable "digital fingerprints" within file system metadata. This approach shifts governance from traditional post-hoc accountability to proactive source control, creating enforceable operational standards. Furthermore, through collaborative oversight, the Regulations impose obligations on content creators, distributors, platforms, and even users in secondary dissemination, forming a closed-loop regulatory framework spanning production, distribution, and redistribution.
《办法》最具突破性的创新，在于构建了“显式+隐式”的双重标识体系。显式标识在生成合成内容或者交互场景界面中以文字、声音、图形等方式强制提醒受众，隐式标识则通过技术措施在生成合成内容文件数据中添加可溯源的“数字指纹”。这种设计跳出了传统“事后追责”的治理逻辑，把责任链延伸到内容生产的源头，为对人工智能的治理提供了可行的操作规范。同时，凭借体系性协同治理，《办法》对内容生产者、分发平台、传播平台甚至用户的二次传播进行约束，形成了从生成、传播到再生产的闭环监管。
Yet challenges to labeling mechanisms are far more complex than anticipated. Technologically, adversarial tactics persist: platforms may evade visible labels, while the black and gray production gang illegally uses watermark-removal tools to tamper with embedded identifiers. Public awareness gaps compound the issue. A Waterloo University study reveals limited public ability to detect AI-generated images (61% accuracy), with emotional bias often overriding critical scrutiny—when even a "Grok AI" watermark in a corner goes unnoticed, labels alone cannot combat synthetic disinformation. Commercial interests further clash with public values: some platforms deliberately delay AI content reviews to exploit the "golden hour" of rumor virality, prioritizing traffic over truth. Behind the convenience of AI lies a shadowy industrial chain, spanning tool development, content fabrication, and monetization.
然而，标识制度面临的挑战远比想象中复杂。一方面，技术对抗持续升级：部分平台采用技术手段规避显式标识的可见性，黑灰产团伙非法利用去水印工具批量篡改隐式标识。另一方面，公众认知存在漏洞。滑铁卢大学的一项研究表明，普通人对AI生成图片的整体识别能力较低（61%），且细节观察的准确率有限，更多人会被情感驱动轻信“有图有真相”。当图片角落的“Grok AI”水印都被集体忽视时，单纯依赖标识显然不足以解决AI合成虚假新闻的问题。商业逻辑与公共价值的冲突也造成了更深层的矛盾。部分平台为争夺流量，故意延迟AI内容审核，利用“谣言传播黄金1小时”收割受众注意力。人们纵享AI便利的同时，“AI污染”已形成从工具开发、内容生产到流量变现的完整黑色产业链。
To truly achieve a "clear and healthy cyberspace," improving public digital media literacy is urgent.When users unquestioningly trust AI-generated images or let fabricated "tears" in videos stir up group emotions, society’s "information immunity" is alarmingly deficient. Combating AI-driven "information pollution" requires not just technological or regulatory fixes but a societal awakening. Without basic discernment skills, even the strictest labeling
systems may fail at the endpoint of dissemination—after all, undetected lies is the same as the truth.
要想真正实现“清朗网络空间”的愿景，培育公众数字媒介素养可谓迫在眉睫。当网络用户对AI生成的图片深信不疑，当伪造视频中的“眼泪”足以煽动群体情绪——在技术滥用面前，公众的“信息免疫力”已严重不足。对抗AI时代的“信息污染”，不能仅靠技术进步或制度约束，更需要一场全民认知觉醒。若缺乏对AI生成内容的基本鉴别能力，再严格的标识制度也可能在传播链末端失效，毕竟，无人识破的谎言与真相无异。
The implementation of the Regulations is not the end but the beginning of this "truth" defense campaign. While China is steering AI toward ethical use through institutional innovation, the Sword of Damocles still hangs overhead: AI governance must perpetually balance innovation incentives with risk control. Only by synchronizing technological norms, social ethics, and legal frameworks can humanity harness AI as a "Promethean fire" illuminating information civilization—rather than unleashing a "Pandora’s box" that devours reality.
在这场“真实”保卫战中，《办法》的施行不是终点，而是起点，我们正以制度创新引导技术向善。但正如古希腊神话中达摩克利斯之剑始终高悬，人工智能的相关治理永远需要在创新激励与风险控制间寻找动态平衡。唯有持续推动技术规则、社会伦理与法律体系的协同进化，才能让人工智能真正成为惠及信息文明的“普罗米修斯之火”，而非吞噬真实的“潘多拉魔盒”。（山东大学 孙政南）