The State of AI-Enabled Malware August 2026: From Brand Abuse to Agentic Execution

This Unit 42 report analyzes 405 AI-themed malware samples and finds that the vast majority (≈97%) are research, proof-of-concept, or security testing artifacts; only 12 samples (~3%) appeared in production telemetry. Observed production samples span ransomware (FunkSec), a trojanized AI application, a backdoor (Oyster), an information stealer (Rhadamanthys), and a COM-hijacking DLL; Palo Alto Networks telemetry shows these were detected and blocked, indicating existing defenses remain effective though AI may accelerate malware development.