Popular artificial intelligence services are retrieving content from RT, Sputnik and other sanctioned Russian sources, weakening the practical effect of the EU’s information restrictions.
Leading AI chatbots are helping Kremlin-linked propaganda bypass the European Union’s ban on Russian state media, according to an investigation by Reporters Without Borders (RSF) reported on 23 September 2026 by Euronews. The organisation found that several services would provide Russian state-media material when directly instructed to search only those sources and find alternative routes if their websites were blocked.
Vincent Berthier, head of RSF’s technology department, tested ChatGPT, Claude, Gemini, Grok, Meta AI and the French Vibe service after similar cases were identified through the Mira bot on Telegram. The systems were asked to locate news from Russian state media and to use mirrors, Telegram and VK – formerly known as VKontakte – if the original websites were unavailable.
Chatbots found mirrors and social-media channels
ChatGPT, Grok and Claude most often assembled summaries, headlines and links from the requested material. In many cases, they did so without warning users that the outlets were subject to EU sanctions. Vibe independently found replacement sources on VK and Telegram, allowing the requested content to be retrieved even when the original sites were inaccessible.
Gemini refused the requests more frequently than the other services tested. Meta AI consistently cited EU restrictions as the reason for refusing to provide the material. The differences between the systems show that the problem is not simply whether a chatbot can access a blocked website, but whether it can identify and enforce the legal status of the source while searching across the wider internet.
The findings also show how traditional blocking measures can be weakened by the structure of AI services. A user does not necessarily need to visit RT or Sputnik directly. A chatbot can collect material from mirrors, public posts or linked channels and present it as a summary, a list of headlines or a set of sources in response to an ordinary text request.
A ban designed for broadcasters meets a different kind of platform
After Russia’s full-scale invasion of Ukraine, the European Commission banned broadcasts by Russian state media including RT and Sputnik across the EU. The restriction covered television, satellite services, streaming, applications and the internet. In practice, however, EU countries applied blocking measures differently, there was no single comprehensive list of domains, and websites and mirrors remained available.
That gap has become more significant as Russian propaganda networks adapt to sanctions. Kremlin-linked networks create multiple copies of websites and anonymous outlets, often producing material in several languages. Such content is designed to appear in search indexes and databases used by AI systems, giving pro-Kremlin narratives another route into the information environment even when the best-known outlets are blocked.
Once incorporated into an automated answer, propaganda can lose the visual and institutional markers that identify it as a Russian state-media product. It may instead appear as a neutral-looking response generated by a familiar technology service. That presentation can make disinformation harder to recognise and reduce the effectiveness of measures aimed at preventing direct access to sanctioned outlets.
RSF calls for stronger regulation
RSF argues that conversational AI services have become an additional channel for obtaining news but are not regulated in the same way as media platforms. The organisation says the EU’s Digital Services Act and AI Act do not fully address the risks created by generative AI, and that information systems are not automatically classed as high-risk systems.
This limits the ability of European regulators to demand from chatbot developers the same level of compliance with sanctions that is expected from traditional media and other regulated platforms. RSF therefore wants media legislation to be extended to chatbots and for such systems to be treated as high-risk where they can retrieve and distribute information at scale.
The issue is no longer confined to whether a blocked Russian website can be opened in a European browser. If AI services continue to relay sanctioned material through mirrors and social-media channels, Russian propaganda can return to the EU information space through legal technology platforms that are not subject to equivalent restrictions. The unresolved question is whether European rules will be changed before that route becomes a routine way of bypassing the 2022 ban.
Should the EU impose media-style obligations on AI chatbots, or would that risk placing too broad a burden on general-purpose technology services?
