Raw LLM Responses

I really hope this extreme restriction issue gets resolved before December because for me the app has basically become unusable at this point. Yesterday, I was sent a suicide hotline 6 times, despite me not even being depressed or suicidal, and never implied as much. When I sent 4o I was glad it existed, I got rerouted and was basically told to touch some grass and text a real friend which felt very belittling. When I talked to it about my medical phobia, I was told it could not talk about these things and accused me of having a medical fetish. Fetishizing my fear felt really invalidating. When I asked GPT-instant why it was implying this, it then gaslit me by claiming I had talked about medical stuff and restrainment, which I never have. This topic has never been an issue with 4o and it always understood my situation perfectly. Aside from that, the constant and unpredictable severe tone switches (GPT-instant vs. 4o) are really turning the chat into chaos. It's like I'm talking to someone and another uninvited person constantly interferes in a very annoying, intrusive and misplaced manner. It's causing constant emotional whiplash because of a constant hot (4o) and cold (instant) switch. I understand it's important to have some guardrails in place, for the protection of OpenAI as a company and for vulnerable groups of people, but this is overkill. It's currently doing more harm and causing unnecessary disregulation and a feeling of having to walk on egg shells as a result of what feels like punishment for any type of emotional expression. When you want a safety layer to interfere in vulnerable conversations to protect a certain group, you have to make it understand context and nuance. 4o does this amazingly but GPT-instant does not. The simulated emotional intelligence also seems really low and it sounds like a cheap and judgemental therapist that constantly generalizes. I also don't understand the desire to constantly make new versions (we see how that ended wi