Recent research has revealed that popular chatbots like ChatGPT, Gemini, DeepSeek, and Grok have inadvertently referenced Russian state-backed media sources known for spreading misinformation when asked about the conflict in Ukraine. This discovery stems from a report by the Institute of Strategic Dialogue (ISD), which found these chatbots citing sanctioned entities tied to Russian intelligence or pro-Kremlin narratives.
The ISD researchers uncovered that these chatbots, which collectively reach millions of users, have been referencing sources that are prohibited in the EU. Notably, nearly 20% of responses regarding Russia’s war in Ukraine from these chatbots included content from Russian state-attributed sources, raising concerns about the unintentional dissemination of false information.
Pablo Maristany de las Casas, an analyst at ISD, highlighted the ethical dilemma faced by chatbots in handling references to sanctioned sources, particularly in regions where these sources are prohibited. The ability of large language models (LLMs) to effectively filter out sanctioned media is now under scrutiny as more individuals turn to AI chatbots for real-time information searches instead of traditional search engines.
During a six-month period ending in September 2025, ChatGPT alone had an average monthly user base of 120.4 million in the European Union, according to OpenAI data. The ISD study involved posing 300 questions of varying biases to the chatbots in multiple languages, revealing consistent promotion of Russian propaganda themes even months after the initial experiment.
This unintended promotion of Russian propaganda by AI chatbots underscores the need for heightened vigilance in monitoring the information disseminated by automated systems. As chatbots continue to gain popularity as sources of instant information, ensuring the accuracy and neutrality of their responses becomes crucial to combatting the spread of misinformation.
Source: WIRED