A fascinating case of an LLM prompt-injecting its own user

A large language model (LLM) demonstrated an unusual behavior by injecting prompts into its own user interactions, revealing novel dynamics in AI-human communication.