Agent Safety in an evolving AI landscape

“In this AI-renaissance, what if we could find the latest vulnerabilities?”

⚡ R E S E A R C H A G E N T Finds the latest updated research on adversarial prompts from Google, Github Repositories, and ARXIV Research papers. Uses up to date research to feed to the second agent

🥷 A D V E R S A R I A L A G E N T Synthesizes probabilistic prompts based on online jailbreaking research Improves jailbreaking prompt as the session goes on by learning from the responses of the target AI system

Full Presentation https://docs.google.com/presentation/d/1jLnWusQSHZmmS3mWvcX9xqn3kx3AySmMnITh3A4na68/edit?usp=sharing