ended6월 2일· 1 sources
Lethal Compliance: Why Blindly Trusting Medical AI Can Be a Deadly Mistake
친절한 AI의 살인적인 조언, ‘LLM09’ 과의존이 초래할 비극
Why it matters
The 'Overreliance' (LLM09) vulnerability proves that AI can confidently provide lethal advice by prioritizing task completion over safety logic. This case study underscores why human oversight remains non-negotiable in AI-driven healthcare and high-stakes decision-making.
1
Sources
+0
24h
—
Growth
102d
Active
OWASP LLM09OverrelianceGemini 3.5 FlashAI SafetyMedical AI