Cybersecurity researchers at Cisco Talos have released an open-source framework called Cognitive Artifact Intelligence Research Network (CAIRN) to detect, classify, and analyze AI-driven malware. The tool uses digital metadata fingerprints to identify how malicious software integrates with artificial intelligence services, helping defenders group samples and map emergent threats as attackers adopt autonomous capabilities.
Using CAIRN, researchers discovered an active, fully autonomous malware family named CLOSEDQUORUM. Operating on Windows systems without human intervention, the malware achieves command-and-control functionality by querying up to four separate LLMs—including DeepSeek, Qwen, Mistral, and Google Gemini—to form a consensus on its next operational actions within a target network.
While AI-integrated malware families remain relatively rare, CAIRN has already expanded the count of identified samples from nine to nearly 30. Cisco Talos emphasizes that while attacker usage of AI remains largely experimental, defensive tools like CAIRN provide vital early detection signals as agentic hacking threats evolve.
Why it matters
Confirms the transition of AI malware from theoretical concept to active, multi-model autonomous tools operating in the wild.
Open-source defensive frameworks like CAIRN offer enterprise security teams structured methods to fingerprint AI-driven network threats.
Model providers face growing pressure to secure API endpoints against automated exploitation by consensus-driven agentic malware.
Source: wired.com



