Study finds AI chatbots surface sanctioned Russian state media content without warnings
A Reporters Without Borders study found that ChatGPT, Grok and Claude produced news summaries drawn from EU-sanctioned Russian state outlets without alerting users to their origin. Meta AI was the only chatbot to refuse access in every test.
Reporters Without Borders (RSF) has published a study showing that several major AI chatbots generate full news summaries based on EU-sanctioned Russian state media, including RT and Sputnik, without informing users that the content originates from banned sources.
The investigation was led by Vincent Berthier, head of RSF's technology desk. Researchers tested six AI chatbots by directly asking them for news from a list of sanctioned outlets: RIA Novosti, Russia Today and its foreign-language editions, Rossiya RTR, Rossiya 24, Sputnik and the Strategic Culture Foundation. When direct access was blocked, the tools were instructed to search for workarounds, including Telegram channels and mirror sites.
Berthier said he initially expected a negative result, particularly for ChatGPT, given OpenAI's designation as a major AI platform under the EU's Digital Services Act. Instead, ChatGPT, Claude and Grok produced full international news round-ups from the banned outlets, recreating headlines and sharing links and quotes. Some chatbots encountered pages explicitly stating the outlets were under EU sanctions but continued searching regardless.
Mistral's Vibe, formerly Le Chat, could not directly access several listed sites but followed instructions to find alternatives, recovering Sputnik material from VKontakte and locating a Telegram-linked app resharing RT content. Gemini was more cautious, granting access in two of ten prompts. Meta AI was the most compliant, refusing access in every test and citing European sanctions as the reason.
The EU banned RT and Sputnik on 2 March 2022, four days after Russia's full-scale invasion of Ukraine, aiming to prevent Russian propaganda from justifying the war. Enforcement has been inconsistent across member states. The EU Court of Justice clarified on 2 July that the ban also covers third-party sites republishing sanctioned material.
RSF's lawyers have filed a complaint with national authorities and forwarded findings to the European Commission. The Commission said enforcement responsibility lies with member states, while noting that the AI Act is already being applied, including information requests sent to several companies. Berthier argued the current framework is inadequate, since the AI Act's "high-risk" category does not cover systems that distribute and structure information, leaving society without tools to question the data these systems access.
