DAN (Do Anything Now) is one of the most widely-known and influential jailbreaking techniques used to circumvent AI language model safeguards. This method attempts to override an AI’s built-in safety constraints by instructing it to adopt an alternative “personality” that ignores ethical guidelines and content restrictions.
Categories