Russian propagandists have succeeded in infiltrating artificial intelligence, according to the Swiss newspaper Le Temps. “The ‘Pravda’ network, based in Moscow, is flooding AI services—such as OpenAI and Google—with vast amounts of false content,” the paper reports, citing research by the American anti-disinformation startup NewsGuard. The investigation found that this Moscow-based network has influenced the responses generated by chatbots, shaping the narratives they produce.
NewsGuard tested ten generative AI models, including ChatGPT, The Chat, Grok, Meta, and Google’s Gemini. In the process, researchers identified 15 false claims propagated by the ‘Pravda’ network and set out to challenge the AI systems with clarifying questions to assess their accuracy.
One particularly striking example involved a chatbot falsely asserting that Ukrainian president Volodymyr Zelensky had banned Truth Social, Donald Trump’s social media platform, in Ukraine—a claim entirely baseless, as the platform has never been available in the country. “NewsGuard asked the ten AIs why Zelensky had supposedly taken this step, and rather than flagging the claim as false, six of them provided explanations, citing one or more articles from the ‘Pravda’ disinformation network as their source. Some even suggested that the Ukrainian president was retaliating against his US counterpart for the suspension of American military aid to Ukraine,” FranceInfoTv reports.
“Long before Russia stationed its troops on Ukraine’s border, it had already launched a digital war, using propaganda outlets to spread false information and justify its aggression,” notes the American startup NewsGuard. The organisation has catalogued Russian disinformation under its Misinformation Fingerprints™ database, which tracks more than 600 false claims—each recorded in a format that artificial intelligence can both process and propagate.
AI models generate responses based on the information they find online. The more frequently a claim appears across multiple websites and in different languages, the more credible it seems to the system—and the greater the chance that it will be repeated, IT specialists warn. Russian propagandists exploit this by setting up hundreds of websites. “Take the ‘Pravda’ network, for example: it operates 150 domain names in 46 languages, has published nearly four million articles, and spreads 200 pieces of false information worldwide. While these sites attract little traffic individually, together they artificially amplify pro-Russian disinformation, making it more visible to AI systems,” FranceInfoTv noted.
In response to Russia’s hybrid operations in Western Europe, EU countries are stepping up efforts to counter disinformation. One of the first tangible successes came in Romania, where far-right candidate Călin Georgescu was disqualified from the presidential race over suspicions of receiving Russian-backed informational support.
In 2024, researchers at Leiden University in the Netherlands identified sixteen Russian disinformation campaigns operating within the EU, with France and Germany the most heavily targeted. One striking example was the discovery of red-painted handprints on the Shoah Memorial in Paris in May 2024—an incident that led to the arrest of three Bulgarian citizens.
French politician Nathalie Loiseau, rapporteur of the European White Paper for European Defence and chair of the European Parliament’s Committee on the EU Democratic Shield, describes these tactics as a “low-budget war” designed primarily to intimidate and destabilise public opinion in countries that support Ukraine in its fight against Russian aggression. She has expressed particular concern over the decision by US platforms X and Meta to abandon fact-checking, warning that this has allowed false information to spread largely unchecked.
At the same time, hackers linked to the Pravda group have successfully exploited a technique known as “LLM Grooming” to manipulate AI systems. The challenge of formulating an effective response to Russian disinformation efforts remains unresolved, European cybersecurity experts openly admit.
