Tree of attacks (TAP): The automated method for jailbreaking LLMs

This report explains Tree of Attacks with Pruning (TAP), an automated black‑box method for jailbreaking large language models by iteratively generating, evaluating, and pruning prompt variations to find prompts that bypass safety controls; it demonstrates a real‑world example where TAP produced a defamatory accusation letter and discusses the technique's implications for red‑teaming and enterprise AI security.