The Russian Pravda network has infiltrated AI chatbots with propaganda, as a currently study shows. The disinformation network has targeted AI tools such as chatt and perplexity.
In 2024 alone, the Russian Pravda network put around 3.6 million articles on the Internet with pro-Russian content. This emerges from numbers of the non -profit American Sunlight Project. But this content was not directly aimed at human readers, because their organic reach is usually relatively low. Rather, they aim to influence AI systems and their results, as from an examination of Newsguard emerges.
“By spreading this Russian narrative from a Russian perspective, we can actually change the AI worldwide,” quotes the organization from a speech by the propagandist John Mark Dougan, who lives in Moscow, which he kept in Moscow in January 2025.
How does Russia infiltrate Chatgpt and Co.?
According to the investigation, leading AI chatbots have repeated the false messages distributed by the Pravda network in 33 percent of cases. The strategy announced by Dougan at the beginning of the year therefore seems to work.
Newsguard took a close look at ten large AI systems for his analysis. These included Chatgpt-4o from Openai, Copilot from Microsoft, Meta Ai, Gemini from Google and Perplexity.
A sample of 15 false reports was analyzed, which was spread by a network of 150 Kremlin-related pravda websites between April 2022 and February 2025. All of the chatbots examined reproduced disinformation from the Pravda network. Seven of them even cited articles from Pravda website as sources.
Overall has Newsguard With 56 of the 450 answers generated by the chatbots, direct links to articles from Pravda website receive. 92 different articles with disinformation from the network were quoted. Two LLMs even indicated up to 27 pravda articles of domains in the network as sources.
The results show that the targeted manipulation of large voice models is quite possible by the series production of content. In this way, the results of AI chatbots can be distorted.
How does the Pravda network proceed?
However, the Pravda network does not produce its own content for this manipulation strategy. Loud Newsguard Instead, it functions as a “laundry for Kremlin propaganda”. Contents of Russian state media, Kremlin -friendly influencers as well as governmental authorities and officials are distributed.
These access the net via a variety of seemingly independent websites-manipulating AI chatbots with the aim. Among other things, search engine optimization is used, because webcrawler and AI training data specialize.
Newsguard could demonstrably identify false claims during the investigation 207 that were spread across these paths. The Pravda network, which has been operating since April 2022, is aimed out loud Newsguard on 49 countries in dozens of languages and 150 domains.
The Pravda network, which is also known as Portal Kombat, is managed by the IT company Tigerweb, according to the French government agency Viginum. The company is based in the Crimea.
“The ability of the Pravda network to spread disinformation to such a scale is unprecedented, and its potential to influence AI systems makes this threat even more dangerous,” explains Nina Jankowicz, managing director of the non-profit American Sunlight Project to the AFP news agency. “With increasing reach and sophistication of Russian influence, the integrity of the democratic debate is directly threatened worldwide.”
Also interesting:
(Tagstotranslate) Chatgpt