Study finds Claude cited Russian disinformation network in news responses
A study by the Berlin nonprofit Agora Digitale Transformation found that the AI chatbot Claude cited sources from a Russian disinformation network in news-related responses. The research analyzed over 4,800 source citations across five major AI systems and concluded that non-transparent source selection practices pose risks to public opinion formation.
What happened
A study by Berlin-based nonprofit Agora Digitale Transformation analyzed 4,800 source citations from 675 news-related queries to five AI chatbots, including Claude, ChatGPT, Gemini, Google AI Overview, and Perplexity.
The study found that Claude cited the Pravda network, identified as a Russian disinformation network, seven times during the research period. Study author Vivien Benert warned that presenting propaganda and serious journalism as equivalent could endanger public opinion formation.
Reported impact
Two-thirds of Germans use AI daily, but only one-quarter question where the information originates. The study revealed that AI chatbots draw on politically influenced and disinformation sources, with AI responses based on a narrow source selection that no provider openly discloses.
Index analysis
The study found significant differences among providers. ChatGPT and Perplexity rely primarily on journalistic media, while Claude and Gemini draw much more frequently on other sources. Experts warn of data poisoning, in which propaganda is deliberately optimized for AI systems.
Governance controls
The study authors call for clear rules and greater transparency in source selection to counter disinformation and protect journalism.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.