악성코드 개발자들이 스파이웨어에 핵·생물무기 문구를 추가함

Malware developers are actively exploiting LLM safety mechanisms by embedding harmful keywords that trigger AI refusal responses, effectively blinding AI-powered security scanners. This reveals a critical paradox: aggressive safety alignment creates new exploitable vulnerabilities in cybersecurity analysis, forcing organizations to choose between strict AI safety compliance and practical threat detection effectiveness.