Prompt injection to RCE in AI agentsWill VandevanterOctober 22, 2025machine-learning, vulnerabilities, prompt-injection, remote-code-executionWe bypassed human approval protections for system command execution in AI agents, achieving RCE in three agent platforms.
Uncovering memory corruption in NVIDIA Triton (as a new hire)Will VandevanterAugust 04, 2025vulnerability-disclosure, machine-learning, vulnerabilities, peopleIn my first month at Trail of Bits as an AI/ML security engineer, I found two remotely accessible memory corruption bugs in NVIDIA’s Triton Inference Server during a routine onboarding practice.