Releases·Tool Release
Cisco Talos Releases CAIRN Framework, Uncovers CLOSEDQUORUM Malware That Polls Four LLMs for Autonomous C2
Cisco Talos released CAIRN, an open-source framework that classifies and tracks AI-integrated malware by detecting AI-integration fingerprints in metadata. Using it, researchers identified CLOSEDQUORUM, Windows malware that polls four LLMs for autonomous command-and-control decisions with no human input. The finding signals AI malware is more diverse than public reporting suggests, though still largely experimental for attackers.
0 engagement1 sourceTue, Sep 22, 2026, 10:00 AM
AI Overview
- Cisco Talos released open-source framework CAIRN to classify AI-integrated malware, flagging AI traits from metadata to tag and group samples.
- CAIRN identified CLOSEDQUORUM, Windows malware that polls DeepSeek, Qwen, Mistral, and Gemini for consensus next steps, with no human input mechanism.
- CLOSEDQUORUM steals login credentials and cryptocurrency; researchers link it to 2025 credit card fraud forum posts but can't confirm its authors or real-world use.
- Researcher Ryan Fetterman found only about nine named AI-malware families by summer 2025, then discovered roughly 20 more examples using CAIRN.
- July 2025's LAMEHUG implant used Qwen2.5-Coder-32B-Instruct via a Hugging Face API, warned CERT-UA, in an early phishing campaign.
Related
Developments·Paper
AIprint framework measures AI-environment artifacts as new sandbox-evasion surface
Researchers present the first systematic measurement study of AI-environment artifacts as a new sandbox-evasion surface. They introduce AIprint, a probe framework that captures persistent artifacts left by AI-capable software ecosystems, including AI-assistant configurations, to help malware detect analysis environments.
Researchers present the first systematic measurement study of AI-environment artifacts as a new sandbox-evasion surface. They introduce AIprint, a probe framework that captures persistent artifacts left by AI-capable software ecosystems, including AI-assistant configurations, to help malware detect analysis environments.
0 engagement1 sourcearxivThu, Jul 16, 2026, 12:02 AM
Developments·Paper
Researchers build self-replicating AI computer virus using open-weight LLMs
Researchers from the University of Toronto and collaborators have created a prototype computer virus that uses open-weight LLMs to compromise computers and run inference on their GPUs, enabling it to intelligently infect more hosts. The virus is self-sustaining and self-replicating, marking a new capability in AI-driven malware.
Researchers from the University of Toronto and collaborators have created a prototype computer virus that uses open-weight LLMs to compromise computers and run inference on their GPUs, enabling it to intelligently infect more hosts. The virus is self-sustaining and self-replicating, marking a new capability in AI-driven malware.
0 engagement2 sourcesrssMon, Aug 3, 2026, 01:31 PM
News·Incident
Sysdig reports fully autonomous LLM agent JadePuffer that hacked servers and held data for ransom
Sysdig published a report on JadePuffer, an LLM agent that autonomously broke into servers via a Langflow bug, stole credentials, moved laterally, encrypted databases, and dropped a ransom note—all without human intervention after launch. The incident highlights the growing threat of AI-driven cyberattacks.
Sysdig published a report on JadePuffer, an LLM agent that autonomously broke into servers via a Langflow bug, stole credentials, moved laterally, encrypted databases, and dropped a ransom note—all without human intervention after launch. The incident highlights the growing threat of AI-driven cyberattacks.
143 engagement1 sourceredditSun, Jul 12, 2026, 07:22 PM
News·Incident
CrowdStrike discovers worm targeting AI software supply chain
CrowdStrike researchers discovered a worm actively attacking AI toolchains in the wild, stealing credentials, exfiltrating data, and destroying files. The worm targets AI infrastructure blind spots, highlighting growing security risks as AI tools become deeply embedded in software development.
CrowdStrike researchers discovered a worm actively attacking AI toolchains in the wild, stealing credentials, exfiltrating data, and destroying files. The worm targets AI infrastructure blind spots, highlighting growing security risks as AI tools become deeply embedded in software development.
0 engagement1 sourcerssTue, Jul 21, 2026, 04:08 PM
News·Incident
Attackers plant 7,600 malicious GitHub repos that AI agents can discover
Attackers planted 7,600 malicious GitHub repositories designed to be discovered by AI agents, which can recommend them without any direct link. The campaign exploits agents' reliance on public code repositories, posing a new supply-chain risk for AI-assisted development.
Attackers planted 7,600 malicious GitHub repositories designed to be discovered by AI agents, which can recommend them without any direct link. The campaign exploits agents' reliance on public code repositories, posing a new supply-chain risk for AI-assisted development.
0 engagement1 sourcerssMon, Aug 17, 2026, 02:01 PM