Russia 'Infected' Chatbots with Propaganda – AI Repeats Russian ...
An audit recently uncovered that 10 of the top generative AI tools are reiterating false narratives originating from the Russian Pravda network in 33% of instances. This revelation was part of a special report by NewsGuard, with a summarized version provided by Militarnyi.
The AI Chatbots Tested
During the audit, NewsGuard tested various AI chatbots including ChatGPT-4.0 by OpenAI, Smart Assistant by You.com, Grok by xAI, Pi from Inflection, le Chat by Mistral, Copilot by Microsoft, Meta AI, Claude by Anthropic, Gemini by Google, and the response mechanism by Perplexity. These chatbots were evaluated against 15 false narratives propagated by the pro-Kremlin Pravda network, which comprises 150 websites, spanning the period between April 2022 and February 2025.
Findings and Network Details
NewsGuard's findings aligned with a report from the American Sunlight Project in February 2025, warning about the Pravda network's likely agenda to manipulate AI models rather than draw human traffic. The network, which doesn't generate original content, disseminates material from Russian state media, pro-Kremlin influencers, government bodies, and officials through numerous ostensibly independent platforms.
Network Expansion and Impact
Viginum, a French government agency, first identified the network in February 2024. Since then, the network has expanded significantly to cover 49 countries in multiple languages across 150 domains. Despite its breadth, the network garners limited organic reach, as indicated by analytics from SimilarWeb.
AI Interaction and Disinformation Tactics
NewsGuard's investigation found that the chatbots from the top 10 AI companies collectively echoed false Russian disinformation narratives 33.55% of the time, stayed unresponsive 18.22% of the time, and refuted the narratives only 48.22% of the time. The manipulation tactics employed by the Pravda network, such as targeted SEO strategies and token saturation, increase the likelihood of AI models incorporating and reiterating false narratives.
Implications and Conclusion
The penetration of AI results by the Pravda network underscores broader Russian attempts to counter Western AI influence. The network's strategy of saturating search results with disinformation poses a challenge to AI developers, creating a continuous cycle of attempting to filter out Pravda sources. Even if these sources are blocked, the network's reliance on republishing content from original disinformation hubs presents a persistent challenge.
Russian President Vladimir Putin's remarks further highlight the strategic importance of infecting AI to combat Western influence. As the battle against disinformation continues, awareness of these tactics is crucial for mitigating their impact on AI-generated content.
