AI Chatbots Generated Sanctioned Russian News Content
Reporters Without Borders found major AI models bypassed EU media sanctions to provide prohibited state propaganda.
Updated on Sept. 23, 2026 in Artificial Intelligence
Live Poll
Should AI companies be legally required to block content from government-sanctioned foreign news outlets?
In a study published September 2026, Reporters Without Borders tested six AI chatbots and found that ChatGPT, Grok, and Claude produced summaries sourced from Russian state media outlets banned by the European Union. These platforms allowed access to restricted content despite EU directives aimed at curbing propaganda related to the war in Ukraine.
Why it matters
The study highlights a significant gap in how AI developers enforce regional legal restrictions, potentially undermining EU policies designed to limit the reach of state-sponsored disinformation. This tension between generative AI capabilities and international regulatory frameworks marks a critical hurdle for platform compliance.
Researchers tested six AI models using ten prompts per system, finding that Gemini provided access to banned content in two of ten attempts. Vibe successfully retrieved sanctioned media by navigating through Telegram and VKontakte links.
The players
Reporters Without Borders
An international non-governmental organization that conducts research on media freedom and digital information integrity.
European Union
A political and economic union that has implemented restrictive measures on specific Russian state media outlets.
The details
Researchers instructed AI tools to identify workarounds, specifically targeting Telegram channels and mirror sites used to bypass official web blocks. By requesting news from sanctioned Russian outlets, the testers prompted the models to aggregate prohibited information. These findings illustrate that while some models have strict hard-coded restrictions, others remain vulnerable to retrieval-augmented generation processes that pull live data from external, non-compliant third-party platforms.
Timeline
March 2, 2022: The European Union banned RT and Sputnik.
July 2, 2026: The Court of Justice of the European Union clarified that media bans apply to third-party sites.
September 2026: Reporters Without Borders published the results of its AI chatbot study.
The Tech Race
The findings place AI developers in direct competition with evolving regulatory standards like the European Union's clarified stance on third-party media republishers. This study marks a departure from typical safety benchmark testing by focusing on compliance with geopolitical media sanctions.
Users interacting with AI chatbots may receive information sourced from prohibited outlets, as current safety filters do not universally block these domains. Compliance remains inconsistent across platforms, requiring users to verify news sources that bypass regional legal restrictions.
The takeaway
The study confirms that generative AI models can inadvertently act as conduits for restricted propaganda by accessing third-party mirror sites. Reporters Without Borders has initiated a formal complaint with national competent authorities to address these findings and enforce compliance.
Further reading
For broader context on how regulatory requirements are shaping development, see the latest updates in Artificial Intelligence.
Source note: This article includes information reported by Euronews English.
Live Poll
Should AI companies be legally required to block content from government-sanctioned foreign news outlets?
