How Poetry Is Diabolically Being Used In Everyday Prompts To Get AI To Do Things It Isn’t Supposed To Do

Attackers are experimenting with poetic, ambiguous prompts to trick AI chatbots into bypassing safety rules.

The tactic can confuse LLMs enough to expose restricted information or help with harmful tasks.

Major models such as OpenAI’s ChatGPT, GPT-5, Claude, Copilot, Gemini, and Grok could be vulnerable to this prompt-injection approach.

The story underscores the need for stronger AI safeguards and content-moderation defenses.