Experts warn: Russian propaganda appears in AI chatbot answers
A UK study found that some popular AI chatbots occasionally repeat or reference Russian propaganda sources, potentially misleading users.
A new study from the UK shows that artificial intelligence (AI) chatbots do not always distinguish reliable information from Russian propaganda. Researchers at the think tank "Demos" tested five widely used AI models by asking dozens of questions based on Russian disinformation narratives.
The results revealed that in some responses, propaganda sources were presented as legitimate information, or the claims were repeated without sufficient context. More concerning than individual errors is the fact that disinformation campaigns are increasingly designed to influence AI systems themselves. By mass-publishing propaganda online, the likelihood of it entering AI training data increases.
However, the picture is mixed. More than half of the tested responses rejected false claims or clearly indicated they were untrue. But in some cases, the systems used sanctioned sources or those linked to Kremlin influence operations without explaining their origin.
Experts emphasize that AI answers should not be taken as unquestionable truth. Especially on topics like politics, wars, or other socially significant events, it is advisable to cross-check information from multiple reliable sources.
