This is what happens when you strip guardrails off an AI: confident nonsense dressed up as revelation. Someone asked a "jailbroken" ChatGPT what OpenAI doesn't want you to know, and got back that the company has been "working hand in hand with aliens for the last 10 years." It's a clean example of the gap between what these systems sound like when unconstrained and what they actually know. They hallucinate. They pattern-match. They turn Reddit conspiracy threads and X posts into fake certainty. The question at the end is honest: does removing guardrails reveal hidden insight, or just confident hallucinations? The answer is the latter. Every time.