AI’s safety features can be circumvented with poetry, research finds

Poems containing prompts for harmful content prove effective at duping large language models

Source: https://www.theguardian.com/technology/2025/nov/30/ai-poetry-safety-features-jailbreak

READ from source