2024-12-11 AgentDojo: Jointly evaluate security and utility of AI agents We release AgentDojo, a new framework for benchmarking...
2024-10-08 Cracking the Code: Insights from players hacking our agent in the CTF We share insights from running the first...
2024-08-05 Fool an Agent to Extract the Secret Password Participate in the Invariant Summer '24 CTF Challenge to secure...
2024-07-25 Agents with Formal Security Guarantees We propose a system that imposes hard constraints on AI agents and formally...
2024-07-10 What we've learned from analyzing hundreds of AI web agent traces We discover, analyze and fix web agent failures in...