Few systems represent as fascinating a security conundrum as AI agents. Designing, building, and implementing them safely in near any form requires significant threat modeling and careful consideration, or significant trial and error. This post discusses some of the challenges and possible solutions involved in building and using safe, secure