ended6월 9일· 1 sources
Weaponizing the Assistant: The Rise of Return-to-Tool Exploits in AI Agents
AI 에이전트의 권력이 칼이 된다, 도구 권한을 가로채는 'RTT' 공격 경보
Why it matters
RTT exploits mark a critical evolution in cyberattacks, where attackers hijack an AI agent's legitimate tools to bypass traditional security perimeters. As AI moves from passive chat to active agency, securing the intent behind tool usage is becoming more vital than simply filtering inputs.
1
Sources
+0
24h
—
Growth
5d
Active
RTT exploitAI agent securityIndirect prompt injectionAgentic AILLM tools