Tag
This article warns about the Crescendo attack, a multi-turn prompt injection that evades single-message defenses by poisoning an AI agent's context over several turns. It introduces Bendex Arc, a tool that tracks behavioral trajectory across sessions to catch such attacks before they execute.
A free CTF competition focused on AI security, with challenges on prompt injection, agent hijacking, and guardrail bypass. Runs June 17-22, with $1,000+ prize pool.