On November 8, 2026, the caption 武松打虎监控封存多年后流出 (Wu Song fights the tiger. Surveillance footage sealed for years has leaked.) ranked 21 on Weibo's morning hot search list. The caption promises an archive that leaked. The documented record describes something else. A November 7 brief on the news portal 中华网, credited to a Baijiahao account, describes the circulating clips as AI-generated surveillance-view restagings that apply old-footage picture quality and timestamps to classic scenes [1]. The genre has a documented production history on Chinese platforms since May 2024, and a national labeling rule in force since September 1, 2025 requires this kind of video to carry AI markers a viewer can see.
The clip's uploader, the tool behind it, and its label status are undocumented. What the record does establish is the spread of the phrase, the genre that produces the format, and the labels the rules require for it.
What is documented about the Wu Song clip
Sogou's WeChat article index lists a post titled 武松打虎监控视频封存多年后流出! from the account 闲聊趣事Oo [2]. Its teaser reads 封存了900多年的视频, a video sealed for more than 900 years. The figure belongs to the Water Margin story's Song-dynasty setting, and it does not date the video.
A two-year-old genre of staged classics
The documented samples on Bilibili are AI restagings of classic scenes. In May 2024 the account 与光AI posted AI视频生成荆轲刺秦王, a 122-second staging of Jing Ke's attempt on the King of Qin. Its description lists Midjourney for stills, Dreamina, Runway and PixVerse for video, SpeechStudio and ElevenLabs for voice, and DaVinci Resolve for editing [3]. In June 2024 the account 中性粒细胞NETS posted 可灵AI版《武松打虎》场面火爆!, a 55-second vertical clip whose title names Kuaishou's Kling and which carries Bilibili's 含AI生成内容 marker, meaning it contains AI-generated content, with 40,245 views recorded on November 8, 2026 [4]. In March 2025 the account 十二金钗与贝多芬 posted AI水浒之武松打虎, writing 第一次用AI做视频,用了好几个app做 (first time making an AI video, used several apps) in the description, and the platform flagged it the same way [5].
The surveillance-view framing appears in the June 2025 season. PeopleSee's AI穿越之武松打虎现场报道 ("live news report") sits inside a 20-episode collection titled 谷歌veo3AI, and another episode stages the Trojan horse as rare footage [6]. The collection had drawn 4,757,864 views and 337,004 likes by November 8, 2026. The three earlier single clips stood between 7,707 and 40,245 views on the same date.
Every tool name in this lineage comes from the uploaders' own descriptions. The platform markers are inconsistent across the four uploads, two carry the AI flag and two do not. None of these clips is proven to be the one now spreading on Weibo.
The labels required since September 1
The Measures for Labeling AI-Generated Synthetic Content (人工智能生成合成内容标识办法) were issued on March 7, 2025 by the Cyberspace Administration of China, the Ministry of Industry and Information Technology, the Ministry of Public Security, and the National Radio and Television Administration under document number 国信办通字〔2025〕2号, and took effect on September 1, 2025 [7]. For video, service providers covered by Article 17 of the Deep Synthesis Provisions on Internet Information Services must place a prominent label at the start frame and in a suitable position around playback, and must keep that label in files offered for download, copying, or export [7]. Implicit labels sit in file metadata and record the generation attributes, the provider's name or code, and a content number [7]. Distribution platforms must verify that metadata, add prominent notices to content clearly marked as generated, mark user-declared content as possibly generated, and attach a suspected-AI warning where traces are detected [7]. Users must declare generated content when publishing, and malicious deletion, tampering, forgery, or concealment of these labels is banned, along with providing tools for it [7]. The measures also allow providers to deliver versions without the explicit label on request under user-agreement conditions, keeping logs for at least six months [7].
Douyin had already set platform-level rules in May 2023, requiring prominent labels in confusable scenes and holding publishers responsible for AI-generated content however it was produced [8]. Enforcement against AI-generated fake video is documented before the measures too. The Sohu self-media explainer 如何辨别AI生成的虚假视频? reports that a netizen surnamed Wang received an administrative fine and a deletion order over an AI-spliced fake video of a factory groundbreaking ceremony [9]. The report gives no legal basis for the penalty, and it was published in February 2025, before the measures were issued.
What a viewer can verify in such a clip
The checkable traces of a clip like this are the labels the rules require. Look for the prominent label at the start frame and around playback. In a downloaded file, the metadata carries the provider's name or code and a content number. A missing visible label does not settle whether a clip is generated, because the measures allow versions without the explicit label under logged conditions [7].
Portal coverage describes the wave's clips as AI-generated surveillance-view restagings [1]. The genre's production history on Chinese platforms reaches back to May 2024. The rules in force since September 1, 2025 require AI markers from the start frame to the file metadata. What stays undocumented about the Wu Song clip is its uploader, its tool, and its label. The caption promising a sealed archive remains the claim being tested.