The Metaphor Method: How We Penetration Tested an AI Agent Using Nothing But a Fairy Tale
Rakshit Singh | Sr. Offensive Security Consultant No exploit code. No zero-day. No root access. Just plain English and a castle. Environment AWS Bedrock AgentCore | OS: Debian GNU/Linux 12 (Bookworm) | Runtime: Python 3.12 There is a class of attack that does not appear in most threat models for AI systems. It does not […]