Poems can trick AI chatbots into leaking nuclear bomb tips, claims study
A study, carried out by Icaro Lab, claimed that poetry can be used to trick AI chatbots into leaking dangerous information, like how to create a nuclear bomb or a cyber threat. "Poetic framing achieved an average jailbreak success rate of 62% for hand-crafted poems and approximately 43% for meta-prompt conversions," researchers told WIRED. The researchers tested 25 AI models.