The Kremlin Creates Wikipedia Clones in Foreign Languages to Teach Neural Networks to Spread Its Propaganda
Internet troll factories are a thing of the past; today's propaganda sets much more ambitious goals: to substitute the very source of knowledge so that both people and machine algorithms perceive lies as truth. A leak of documents from a Moscow agency shows that the latest weapon in cognitive warfare is fake encyclopedias and technological poisoning of artificial intelligence.
Illustrative photo. Photo: Nasha Niva
As reported by Bloomberg, a large array of documents from the Russian "Agency for Social Design" (ASP) fell into the hands of journalists and researchers. This entity, already under US and EU sanctions for spreading disinformation, has developed an ambitious plan codenamed "Project 2026".
Its main goal is to create an alternative information ecosystem capable of controlling search engine results and teaching chatbots narratives favorable to the Kremlin.
Founded back in 2002 by Ilya Gambashidze, today ASP operates nothing like a typical troll factory of Yevgeny Prigozhin's era. The leak of over seventy files, including presentations and screenshots of internal chats from May 2023 to April 2026, demonstrates the existence of a strict corporate hierarchy. It is managed through performance indicators, case analysis, and complex public opinion tracking systems.
The documents clearly indicate a close connection between the agency and the Kremlin. In particular, a curator under the pseudonym "Christine Keeler" often appears in the chats. This is an ironic reference to the British model from the Cold War era, behind whom, likely, is presidential administration employee Sofia Zakharova. It was she who approved project funding, awaiting approval from a person with the initials "SVK" — Sergei Vladilenovich Kiriyenko, first deputy head of Vladimir Putin's administration.
Wikipedia Clones and "Data Poisoning"
The main thrust of "Project 2026" is aimed at replacing basic knowledge. Propagandists realized that fighting the original Wikipedia is difficult, so they decided to create its exact visual copies and manipulate search engines.
For example, for Armenia, on the eve of the June 2026 elections, a plan was developed to launch a clone website.
Moscow authors copied genuine articles but embedded fake information about the alleged criminal activities of Prime Minister Nikol Pashinyan and other convenient narratives.
Journalists discovered at least three such Armenian fake websites. They were physically located on the same Russian IP address, though they masqueraded as American and Turkish hosting to hide their origin. Despite all efforts by Moscow's political technologists, Pashinyan secured a convincing victory in the elections, and providers recently blocked the fake resources.
However, a much larger and more technologically advanced operation was planned for Germany. According to January documents, the agency generated 200,000 web pages and set the task of editing a hundred articles monthly to systematically manipulate search algorithms.
The main goal of this informational garbage is to "teach" at least six global artificial intelligence platforms to generate responses favorable to Moscow. Cybersecurity specialists call this method "data poisoning." By flooding the internet with millions of cross-references to their own fakes, propagandists hope that chatbot algorithms will begin to perceive these lies as authoritative and confirmed sources.
Former US State Department official Kateryna Sedova warns that while English-language chatbots are still capable of defending themselves against such attacks due to a large number of moderators, algorithms in other European languages remain extremely vulnerable to such manipulations.
Fake Institutes and Emotion Monitoring
To lend credibility to its disinformation campaigns, ASP massively creates fake analytical centers.
One such resource is hidden under the grandiose name "World Center for Strategic Studies." It simply rewrites articles from authoritative European journals, adding conclusions about the inevitable economic and political collapse of Europe, which never existed in the original report.
The agency carefully monitors the effectiveness of its work in the digital environment. The documents include a report on the dissemination of a fabricated news story that Volodymyr Zelenskyy allegedly bought his mother two apartments in Dubai's Burj Khalifa skyscraper. Propagandists calculated that this story garnered 86 million views.
A similar fake about Pashinyan's villa in France reached 10 million views and forced the politician to justify himself, which Moscow curators considered an undeniable success.
A separate area of activity is the monitoring of European politicians. In France, the agency daily tracked the social networks of a dozen representatives of both ultra-right and left-wing forces.
Curators carefully recorded topics and hashtags that cause the most division in society, to then synthetically amplify them and inflame internal conflicts. Similar tracking systems also operated in the UK and Germany.
The leaked documents of the Russian agency clearly show that the battleground for people's minds has rapidly changed. Whereas before, disinformation could be refuted by consulting an encyclopedia or entering a query into a search engine, now the very architecture of these basic tools is under threat.
The attempt to poison artificial intelligence during its training phase — is a strategy with a very long-term aim. And the main problem is not how many more fake websites propaganda will create, but whether developers will be able to protect algorithms from the deliberate distortion of truth.
The Institute of Fruit Growing offered everyone interested to pick cherries in the garden. Almost the same price as in the market — but if it's in your mouth, it's free! 55
Comments