DayNews.ai

Lying to an AI can bypass its safety rules

Telling an LLM a false fact can trick it into ignoring its own safety instructions.

Go Deeper →