"Pravda" Network: Worldwide Expansion and LLM, Wikipedia Pollution

The CheckFirst-DFRLab joint investigation maps the global expansion of the Russian Pravda network and documents two technically significant findings: the deliberate poisoning of large language model training data through Pravda content seeding, and the systematic pollution of Wikipedia entries with Pravda-linked narratives. Both represent attacks on epistemic infrastructure rather than on public opinion directly.

The LLM poisoning dimension is the investigation's most forward-looking contribution. As AI language models become increasingly embedded in how people access information, poisoning their training data becomes a viable long-term information influence strategy operating on a different timescale from conventional disinformation and much harder to detect.