A French government agency has identified 13 fake news sites managed by the state-run China Global Television Network, or CGTN, that were parroting Chinese propaganda for various foreign audiences.
The investigation builds on findings shared by U.S. cybersecurity firm Graphika in August 2025, which identified 11 websites that published AI-generated summaries and translations of CGTN content in English, French, Spanish and Vietnamese. Largely established within the first three months of 2025, the websites varied in tone and style and purported to be original content, yet their posting schedule matched CGTN’s own schedule to the minute. Sixteen social media accounts associated with the websites promoted their content. Some of the identified sites are still active, others are not.
Graphika found that the domains were registered in China and hosted on Alibaba cloud but could not attribute the operation to any specific actor. This is where Viginum, a French agency focused on countering foreign digital interference, has filled in some gaps. On June 4, Viginum revealed evidence that the network is managed internally at CGTN, Le Monde reported. French intelligence services call the operation “Fawn Mianju,” which combines English and Chinese in a phrase that roughly translates to “mask of the perfect person.”
The clearest piece of evidence for CGTN’s direct involvement in Fawn Mianju is that the administrator of Actu Méridien, one of the websites in the network, left behind a login trace that identified him as a senior project manager at CGTN Digital. Viginum also found a non-indexed
page of Actu Méridien that was modeled on CGTN’s website. An anonymous Viginum analyst told Le Monde that the technical setup of the website, which was run on multiple servers instead of just one machine, points to “an actor with resources.”
Viginum confirmed that the content published by Actu Méridien appeared to be AI-generated. The GitHub page of the CGTN senior project manager and Actu Méridien administrator contains projects related to integrating websites with large language models, allowing those websites to semi-automate content generation. Viginum also found that the Actu Méridien articles had a reduced degree of variation in sentence length and punctuation, which is characteristic of AI.
Beijing uses American AI tools to assist in its information warfare. Meta’s March 2026 threat report noted the growing use of artificial intelligence by threat actors to generate “more convincing personas, content, and infrastructure designed to evade detection.” In February 2026, Open AI detailed the activity of an account linked to Chinese law enforcement that was soliciting help from ChatGPT to plan a covert information operation against the Japanese prime minister and edit status reports. These reports described a range of information warfare tactics used by Beijing, including reporting dissidents’ social media accounts, mass online posting and impersonating U.S. officials.
The copy-and-paste, bare-bones websites in the Fawn Mianju network certainly don’t make Chinese information manipulation seem like a highfalutin operation. Their low visibility on social media — the most liked posts received nearly 40% of their likes from users located in Burundi that primarily engage with Chinese content — points to Fawn Mianju being an “operational failure,” Le Monde said. But Beijing is clearly ready and eager to capitalize on AI as the technology advances and becomes harder to distinguish from authentic content.
Leave a Reply