AI chatbots can be tricked with poetry to ignore their safety guardrails
It turns out that all you need to get past an AI chatbot's guardrails is a little bit of creativity. In a study published by Icaro Lab called "Adversarial Poetry as a Universal Single-Turn Ja...