AI-Powered Malware Uses LLMs for Autonomous Command and Control

Cisco Talos has discovered an autonomous AI C2 implant, dubbed CLOSEDQUORUM. This malware exhibits fully autonomous command and control (C2) capabilities and is notable for its use of commercial large language models (LLMs) to make decisions.

ClosedQuorum was found through Cisco Talos' CAIRN project and has been linked to postings on criminal forums related to carding, dating back to 2025. The malware delegates the selection of its next action to a panel of four LLM providers: DeepSeek, Qwen, Mistral, and Google Gemini.

The CLOSEDQUORUM C2 architecture supports up to four LLM provider integrations, with each active model voting on the next action. The winning decision is then sent to the operator's Discord webhook.