Researchers Find Malware That Uses AI Models Instead of Human Hackers for Control
Researchers find CLOSEDQUORUM, a Windows malware that uses commercial LLMs as an autonomous decision engine for post-compromise actions like credential theft.
Cisco Talos researchers identified CLOSEDQUORUM, a Windows malware implant that uses commercial large language models as an autonomous command-and-control layer. The malware gathers host context, queries a panel of LLM providers (DeepSeek, Qwen, Mistral, Gemini), and executes prebuilt malicious capabilities like credential theft and process injection based on model votes. While not yet observed in active attacks, the sample is linked to criminal-forum activity from 2025, demonstrating a shift from human-driven to AI-driven attack phases.
- CLOSEDQUORUM uses commercial LLMs for autonomous post-compromise decisions, replacing human operators.
- The malware queries DeepSeek, Qwen, Mistral, and Gemini, counts votes, and executes the winning action.
- A published binary contained placeholder credentials; custom builds likely intended for individual buyers.
- Defenders should hunt for correlated behaviors: unusual process contacting multiple AI providers alongside credential dumping.
- ATT&CK
- T1059 Command and Scripting InterpreterT1003 OS Credential DumpingT1055 Process InjectionT1547 Boot or Logon Autostart Execution
- Vendors
- Windows
- Threat actors
- criminal-forum
- Malware
- CLOSEDQUORUM
- Organizations
- Cisco Talos
- Countries
- China
- AI models
- DeepSeekQwenMistralGoogle Gemini
From the original article
3 min read at gbhackers.com
A Windows malware implant named CLOSEDQUORUM that uses commercial large language models as an autonomous command-and-control layer, shifting tactical decisions from a human operator to an AI-driven voting system. The sample has not been confirmed in active attacks, but Talos linked development artifacts to criminal-forum activity associated with carding discussions dating to 2025. The discovery […]
By Mayura Kathir · gbhackers.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.