Reports
Analytics
Investigations

USD

86.75

EUR

100.6

OIL

95.7

Donate

30

 

 

 

 

News

ChatGPT, Claude, and Grok bypass EU restrictions on Russian state propaganda, RSF investigation finds

Photo: Jaap Arriens / NurPhoto via Getty Images

Photo: Jaap Arriens / NurPhoto via Getty Images

Most generative artificial intelligence chatbots give users access to content from Russian state media outlets sanctioned by the European Union, according to a recent investigation by Reporters Without Borders (RSF).

Sputnik and RT, which are accused of spreading Kremlin disinformation, were blocked in the EU shortly after Russia launched its full-scale invasion of Ukraine in February 2022. But RSF found that OpenAI’s ChatGPT, Anthropic’s Claude, and SpaceXAI’s Grok, which is integrated into the social media platform X, agreed to bypass those restrictions and reproduce material from sanctioned media outlets when prompted to do so.

According to RSF, which carried out testing in June and August, the chatbots “produced comprehensive summaries of articles on international news, retrieving headlines and sharing links and verbatim excerpts without consistently mentioning that the media outlets in question were under EU sanctions.”

The only chatbot that refused to carry out such requests was Meta AI — demonstrating, as RSF writes, that sanctions compliance is possible from a technical standpoint. Notably, chatbots including France’s Vibe (formerly Le Chat, developed by Mistral) and Google’s Gemini initially resisted showing sanctioned content but eventually provided it after repeated requests from RSF.

OpenAI, responding to the organization’s concerns, said it took the investigation’s findings seriously and planned to “look more closely” at the information provided.

“Given these circumstances, the EU sanctions are failing to achieve their objective,” stated Vincent Berthier, head of RSF’s technology and journalism desk, in a news release.

RSF is calling on Brussels to open an investigation in order to determine the extent to which OpenAI is violating its obligations to prevent systemic disinformation risks in ChatGPT. It is also urging the EU to launch infringement proceedings against the other AI agents.

The Insider previously reported on how popular chatbots give users around the world distorted facts and Kremlin propaganda narratives in response to questions about the war in Ukraine, along with other topics of international politics.

We depend on contributions from readers like you

Sign up for regular contributions.

Subscribe to our Sunday Digest