For years, cybersecurity professionals have relied on digital fingerprints to track malware and detect infections. However, the rise of agentic AI components in hacking tools demands a new approach. ### Introducing CAIRN: Tracking the AI Malware Frontier **Cisco Talos** researchers, led by **Ryan Fetterman**, have developed the **Cognitive Artifact Intelligence Research Network (CAIRN)**. Named after the stacks of stones hikers use to mark trails, **CAIRN** is an open-source framework designed to identify and classify malware that integrates AI services. "The core idea is that AI integration has these vestiges, like fingerprints, that are left behind," explains Fetterman. "That gives us a signal that we can use to track these samples, classify them, and look at what's happening. What are attackers trying? What kind of emergent behaviors are we seeing? That's a valuable resource to the defensive community as these things become more mainstream." ### CLOSEDQUORUM: The First Autonomous AI C2 Implant Using **CAIRN**, **Cisco Talos** has already identified a hacking tool with fully autonomous command-and-control (C2) infrastructure, dubbed **CLOSEDQUORUM**. This malware plots its actions within a target system by polling up to four large language models (LLMs) – **DeepSeek**, **Qwen**, **Mistral**, and **Google Gemini** – to determine its next steps, effectively creating a hive mind for directives. **CLOSEDQUORUM** is Windows malware designed to steal login credentials and cryptocurrency. Researchers observed links between this malware and cybercriminal forums discussing credit card fraud, dating back to 2025. What makes **CLOSEDQUORUM** particularly notable is its closed-loop system: even if one AI service is unavailable, the malware continues to poll others, ensuring redundancy and eliminating the need for human input. ### The Evolving Landscape of AI-Integrated Malware The development of **CAIRN** was prompted by an observation following a 2025 warning from the Ukrainian cybersecurity response unit **CERT-UA**. They identified a phishing campaign using malware called **LAMEHUG**, which communicated with the **Qwen2.5-Coder-32B-Instruct** LLM via a **Hugging Face** API for commands. Initially, Fetterman anticipated a rapid proliferation of AI-enabled malware. However, a retrospective review revealed only a handful of documented examples, many of which were proofs of concept. This discrepancy spurred the creation of **CAIRN**. "I think I also had a hard time believing that that was the reality of where we were. So I wanted to start digging into that," Fetterman stated. Through **CAIRN**, he has since uncovered approximately 20 additional examples of AI-integrated malware. "So while I do think this is still largely experimental for attackers, the landscape is a lot more complex and diverse than has been publicly reported," Fetterman noted. "There's a lot going on out there, and it does provide a valuable early signal to what's going to happen." ### Operationalizing AI for Attackers **Matt Olney**, Senior Director of Threat Intelligence at **Cisco Talos**, highlights the shift in AI's role. "Initially, everyone saw AI as a productivity tool, right?" he says, referring to both legitimate work and malicious hacking. "Now what we're seeing is that it's becoming operationalized. So for attackers, it's allowing them to run more campaigns, hit more spaces, handle more and different computers, because they have this very intelligent box in the backend that can ask questions and give responses." **CAIRN** is designed to flag AI-integration characteristics and attributes from metadata, classifying and tagging malware samples with unique IDs. The system then analyzes these artifacts within the **CAIRN** library, grouping them by various traits to identify potential trends and connections. This open-source framework offers a critical tool for the defensive community to stay ahead of the rapidly evolving AI-driven threat landscape.