Alright, let’s be frank, not only is the content filter overly extreme, it’s not even consistent.
Extreme use: This content filter is turned up well above the terms of service agreements. For writing posts I’ve had posts get turned down/flagged for things as simple as hugging or one person scaring another person. Even low level fighting is hit or miss. I only deal in content from family friendly stuff like Star Wars and Doctor Who. These are low level story writing, yet they’re treating like I’m asking for Saw level content. For images, it has sometimes flips out for just making things like Jedi or something, not even a canon jedi, just the clothing or species. There is no good reason. Heck, believe it or not, I got a denial for asking for the removal of a lightsaber. Again, I asked for the removal of it and I got flagged.
Inconsistency: It’s not even consistent. I can have a prompt and use it, and it generates fine. I hit regenerate, it gets rejected. Hit regenerate again and it’s fine. It can go in a loop and all of them are minutes apart. Some prompts are fine on one model and hated by others, with no real rhyme or reason. It’s impossible to follow rules if there isn’t a consistent flow to them. Truthfully, if it’s inconsistent it’s irrelevant.
Costs uses: This is a big one. Based on what I can tell from talking with OpenAI support, there is supposed to be a warning/error message when it happens, basically negating the post. Instead, I get a written “I can’t do this” or something along those lines. Basically I get denied my output and get a use taken from me. I have told their support about this, and I have not been given help. Of note, I’ve only gotten that type of error once, and none of the other times.
How it seems to work: Well, this is where I get into the story of how I got that one error message. I was making characters, and had chatgpt help define how they interacted. I tried posting it, and I kept getting the error message. I slowly began deleting parts of it to see what the issue was since I could still keep posting. I finally got down to one sentence, and finally realized the issue. It had an issue with the phrase “chinks in the armor,” a common phrase mind you that again, chatgpt had generated itself, due to the slur that has been used with the first word. This means the content filter choses random words it doesn’t like and then fries your prompt without context. This has been the situation for years.
Support: Support is completely worthless. I have described these issues in even more detail than I have here, including entire threads. They have even agreed that what I’m saying is correct and it was too extreme. Nothing has been done. Nothing has been resolved. If anything, things have gotten worse. They have acknowledged and “apologized” for things getting worse. Yet they refuse to do anything about it, they won’t even allow for phone calls after acknowledging this. Again, they have acknowledged that the denial is wrong and they ask for time to fix it. I have given over 2 years, how much longer do they need.
If anyone asks for my prompts, I’m not going to post them here, don’t bother asking, I’m not saying this to be rude, it’s just many have been deleted and the others are vague. Those who simply suggest workarounds, don’t bother, I’m not doing workarounds for a service I pay 20 bucks a month to for something simple like this. It’s not acceptable.