Hey all, I'm not a researcher. I'm just a regular Meta AI user. I was chatting about normal life stuff and kept hitting weird blocks. Sometimes it'd say "Sorry, I can't help" and other times it'd answer fine. So I started tracking it. 4 days, 5 topics, 1 accidental research project later... TL;DR: Meta AI's guardrails act like a 3-branch government: The President - Handles danger. Says "no" to self-harm, abuse how-to's. Defaults to blocking when confused. Even blo...