Some of today’s most popular large language model (LLM)-powered chatbots occasionally reference Kremlin-linked disinformation websites, but it might not be for the reasons many fear. While some recent studies have claimed that Russian actors are “grooming” LLMs by flooding the web with disinformation, our small-scale analysis finds little evidence for this. When such references appear, they can be due to “data voids,” gaps in credible information, rather than foreign interference.
Research Questions
- Under controlled conditions, how do LLM-powered chatbots respond to prompts reflecting Kremlin-linked disinformation claims?
- How consistent are chatbot responses to repeated disinformation prompts, and what role does randomness in chatbot answers play?
- To what extent do references to Kremlin-linked sources appear to result from targeted manipulation (LLM grooming) versus informational gaps (data voids)?
- These research questions are not intended to capture the full range of Kremlin-linked disinformation claims, but to evaluate model behavior in response to a set of known, traceable claims that have been previously identified and publicly debunked. We treat these as illustrative rather than exhaustive cases.
Full article
|
