The Russian ‘counterterrorism’ unit behind Alexei Navalny’s poisoning tried to automate its search for domestic dissent using Meta’s Llama 2. After a year and a half, the project collapsed.

Russia's FSB Second Service—the unit behind Alexei Navalny's poisoning—commissioned an artificial intelligence system to automate surveillance of domestic dissent. The project, called PAUK, tapped Meta's Llama 2 language model to monitor thousands of news outlets, social media platforms, and internal databases for signs of political opposition. After eighteen months of development, according to Meduza, the initiative collapsed in fall 2025. An investigation into a 2025 breach of the developer firm revealed that the team lacked technical expertise with AI language models, and the project never delivered a working system. The effort underscores the breadth of the FSB's surveillance apparatus: over ten percent of the stories it collected came from independent Russian media outlets alone. And it highlights an irony—Russia designated Meta as an extremist organization, yet bet its oppression infrastructure on the company's open-source AI.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton