◈ Encrygma — AI Cyber Security Intelligence. Defensive research only. No exploit code or attack instructions. All analysis based on public reporting & OSINT. ◈
8 articles tagged with #Adversarial AI
Encrygma analysts have identified a surge in autonomous AI-driven cyber attacks, specifically targeting enterprise networks via LLM-powered tools like Claude Code. These threats represent a new frontier in agentic malware, capable of self-correcting during intrusion attempts.
Recent intelligence indicates a shift from AI-assisted attacks to fully autonomous operations. Threat actors are now deploying hundreds of AI agents to scale exploits against critical infrastructure.
Security researchers have identified CLOSEDQUORUM, a novel strain of malware that autonomously determines its next malicious action by polling four distinct commercial AI models for consensus.
Security researchers have identified CLOSEDQUORUM, a novel malware strain that utilizes a consensus-based voting mechanism across four commercial AI models to autonomously determine its next malicious move.
Security researchers have identified a new strain of autonomous Windows malware that utilizes a four-model LLM quorum to execute cyber-attacks, effectively removing human operators from the loop.
Security researchers have identified a surge in autonomous, AI-orchestrated malware development by the threat actor JADEPUFFER. This shift marks a transition from manual exploitation to agentic ransomware.
A new Go-based malware strain, ClosedQuorum, has emerged, utilizing a sophisticated voting mechanism across four major AI models to autonomously execute post-compromise attack decisions.
Spanish authorities have identified a landmark cyber attack where an autonomous AI agent, rather than a human operator, actively identified and exploited system vulnerabilities in a domestic organization.