Open dataset: Kremlin-linked sources in AI chatbot answers
Dear colleagues, My name is Pavel, I lead Dosimeter, a nonprofit in Berlin, and we monitor Kremlin disinformation in 10 popular AI chatbots, from DeepSeek to Grok. Our Berlin team conducts such measurements on AI services regularly and publishes all this data in the AI Disinformation Benchmark <https://dosimeteraudit.org/benchmark>, making the whole dataset available to researchers and journalists. The data is on GitHub <https://github.com/gurovdigital/ai-disinformation-benchmark> and Hugging Face <https://huggingface.co/datasets/DosimeterAudit/ai-disinformation-benchmark> . Note: In determining what qualifies as disinformation, we relied on databases from EUvsDisinfo, CheckFirst, Detector Media, and other fact-checking organizations. Maybe you find it interesting. Best regards, Pavel Gurov dosimeteraudit.org
participants (1)
-
Pavel Gurov