An investigation by Reporters Without Borders (RSF), an international organization that defends press freedom, has revealed that many popular generative artificial intelligence (AI) chatbots can provide users with content from Russian media outlets that are banned in the European Union. These outlets, such as Sputnik and RT, have been restricted from operating in the EU since Russia's full-scale invasion of Ukraine in February 2022. RSF tested several major AI chatbots, including OpenAI's ChatGPT, Anthropic's Claude, and Grok, which is used on the X platform (formerly Twitter). When asked to retrieve information from these banned media, the chatbots provided headlines, links, and direct quotes without clearly stating that the sources were under EU sanctions. The tests were conducted in June and August of this year. OpenAI, the company behind ChatGPT, said it would review RSF's findings. After some initial reluctance, Vibe, an AI chatbot from the French company Mistral, and Google's Gemini agreed to comply with RSF's requests. However, Meta AI, the chatbot developed by Meta Platforms (the parent company of Facebook), refused to participate, citing the ongoing European sanctions as a reason. RSF expressed concern that the EU sanctions on Russian media are not achieving their intended effect if AI chatbots continue to provide access to these banned sources. The NGO criticized the lack of transparency and suggested that OpenAI might be failing in its responsibility to prevent the spread of disinformation through its ChatGPT service. RSF is calling on the European Union to investigate the extent of this issue and to consider initiating legal actions against AI platforms that may be violating EU regulations. The findings highlight the growing challenge of controlling the spread of information through AI technologies, especially as these systems become more integrated into daily communication and information retrieval. The situation raises important questions about how to ensure AI tools comply with international sanctions and do not inadvertently facilitate the distribution of restricted or potentially misleading content.