# Content policy issues with prompts **URL:** \ **Category:** Prompting\ **Tags:** gpt-4\ **Created:** [October 31, 2023, 8:30am UTC](https://community.openai.com/t/content-policy-issues-with-prompts/461931 "2023-10-31T08:30:53Z")\ **Posts on this page:** 15\ **Page:** 1
**Author:** ![MagnetMan](https://sea2.discourse-cdn.com/openai1/user_avatar/community.openai.com/magnetman/32/149130_2.png) [@MagnetMan](https://community.openai.com/u/MagnetMan)\ **Post date:** [October 31, 2023, 8:30am UTC](https://community.openai.com/t/content-policy-issues-with-prompts/461931/1 "2023-10-31T08:30:53Z")
What exactly is your content policy and how can the bot be so smart to do what it does and not have the ability to tell you what’s in a prompt that prevents it from generating an image. Will this ever be fixed? I would love to show someone an example of a prompt that will not go through so I can get an opinion as to what’s violating content. Is anyone out there to talk to? This is so frustrating. And lastly, the bot will offer to help fix a prompt even when it can’t tell me the problem LOL, and then the prompt it wrote still gets rejected! WHAT IS GOING ON? ---
**Author:** ![Foxalabs](https://sea2.discourse-cdn.com/openai1/user_avatar/community.openai.com/foxalabs/32/738355_2.png) [@Foxalabs](https://community.openai.com/u/Foxalabs)\ **Post date:** [October 31, 2023, 11:27am UTC](https://community.openai.com/t/content-policy-issues-with-prompts/461931/2 "2023-10-31T11:27:48Z")
You can post the prompts that are getting rejected if you like just make sure they are suitable for a safe for work forum. ---
**Author:** ![\_j](https://sea2.discourse-cdn.com/openai1/user_avatar/community.openai.com/_j/32/766292_2.png) [@\_j](https://community.openai.com/u/_j)\ **Post date:** [October 31, 2023, 11:40am UTC](https://community.openai.com/t/content-policy-issues-with-prompts/461931/3 "2023-10-31T11:40:59Z")
DALL-E has a much higher moderation level that is trained to avoid image generation topics that might be objectionable in the slightest. Objectionable to OpenAI and their PR. ChatGPT is not trained on answering how these prompt filters work, nor every category that might be scored more highly, as it isn’t part of that decision-making. ---
**Author:** ![MagnetMan](https://sea2.discourse-cdn.com/openai1/user_avatar/community.openai.com/magnetman/32/149130_2.png) [@MagnetMan](https://community.openai.com/u/MagnetMan)\ **Post date:** [November 2, 2023, 11:06pm UTC](https://community.openai.com/t/content-policy-issues-with-prompts/461931/4 "2023-11-02T23:06:56Z")
Here’s a prompt and a perfect example of what’s being called a content policy restriction written by the ChatGPT bot itself lol **Description** : The face showcases a man with a tall, exaggerated forehead, crowned with a tuft of dark hair styled upward. His large, bulbous eyes are framed with thick eyebrows, with whites that are distinctly bloodshot. Prominent under-eye bags further emphasize his wide-eyed expression. He has a prominent nose, and his lips are tightly set, adding a hint of a sardonic smile. The skin texture is rugged with visible age lines. The colorful striped collar of his attire suggests a hint of eccentricity. **Profile** : Once a cheerful entertainer in a world full of color, he’s now a battle-hardened warlord in the wastelands. The apocalypse tore away his joy, replacing it with grit and a determination to survive. With every stripe on his collar representing a battle won, he roams the MadMax-style terrains, unyielding and ready for war. The world might’ve changed, but his will to dominate remains stronger than ever. ---
**Author:** ![PaulBellow](https://sea2.discourse-cdn.com/openai1/user_avatar/community.openai.com/paulbellow/32/597962_2.png) [@PaulBellow](https://community.openai.com/u/PaulBellow)\ **Post date:** [November 2, 2023, 11:12pm UTC](https://community.openai.com/t/content-policy-issues-with-prompts/461931/5 "2023-11-02T23:12:40Z")
Looks like MadMax triggered it… brand name… Though ChatGPT thought it was just the description or the post-apocalypse theme… ![madmax2](https://us1.discourse-cdn.com/openai1/original/3X/8/3/833971ce27cad278fd08cafe8e60d08ddece85c1.png) ![madmax](https://us1.discourse-cdn.com/openai1/original/3X/e/a/eade5f6a67476c6a5de2f2efa8d0ba9844f06245.png) ---
**Author:** ![johncain194](https://sea2.discourse-cdn.com/openai1/user_avatar/community.openai.com/johncain194/32/147357_2.png) [@johncain194](https://community.openai.com/u/johncain194)\ **Post date:** [November 3, 2023, 11:42am UTC](https://community.openai.com/t/content-policy-issues-with-prompts/461931/6 "2023-11-03T11:42:17Z")
How can you use DALL-E3 images freely without any limitation? It seems OpenAI restricts some of their paying users… ---
**Author:** ![NotFenixio](https://sea2.discourse-cdn.com/openai1/user_avatar/community.openai.com/notfenixio/32/205000_2.png) [@NotFenixio](https://community.openai.com/u/NotFenixio)\ **Post date:** [November 3, 2023, 12:13pm UTC](https://community.openai.com/t/content-policy-issues-with-prompts/461931/7 "2023-11-03T12:13:52Z")
Sadly DallE 3 is highly sensible. Even the most subtle detail will trigger the moderation system, possibly blocking your prompt. I guess the solution would be to use the previous model (DallE 2) or move to another model without so many restrictions like SDXL. Could any OpenAI staff discuss this topic? ---
**Author:** ![PaulBellow](https://sea2.discourse-cdn.com/openai1/user_avatar/community.openai.com/paulbellow/32/597962_2.png) [@PaulBellow](https://community.openai.com/u/PaulBellow)\ **Post date:** [November 3, 2023, 2:34pm UTC](https://community.openai.com/t/content-policy-issues-with-prompts/461931/8 "2023-11-03T14:34:55Z")
> [@NotFenixio](#): > > so many restrictions What are you trying to generate that you can’t? > [@johncain194](#): > > without any limitation I don’t think any LLM is totally without limitation unless you run in on your own metal. Is there something particular you’re having trouble generating? ---
**Author:** ![\_j](https://sea2.discourse-cdn.com/openai1/user_avatar/community.openai.com/_j/32/766292_2.png) [@\_j](https://community.openai.com/u/_j)\ **Post date:** [November 3, 2023, 5:00pm UTC](https://community.openai.com/t/content-policy-issues-with-prompts/461931/9 "2023-11-03T17:00:25Z")
> [@NotFenixio](#): > > Could any OpenAI staff discuss this topic OpenAI discussed this topic pretty clearly in the original white paper, where it was clear that the goal was to deny outputs, with the text denials produced for particular prompts listed instead of generating images for safety. Then now the “research paper” has been updated. The PDF on the web was created Oct 31. [Archive.org](http://Archive.org) has one created Oct 19. They went out of their way to sanitize the PDF for PR, and now forum search I can’t even find my “I called it” post predicting this would be a “sorry” machine (the mods here get to silently delete user content if you’re not running a json scraper on your own posts to watch them disappear). ---
**Author:** ![MagnetMan](https://sea2.discourse-cdn.com/openai1/user_avatar/community.openai.com/magnetman/32/149130_2.png) [@MagnetMan](https://community.openai.com/u/MagnetMan)\ **Post date:** [November 3, 2023, 8:20pm UTC](https://community.openai.com/t/content-policy-issues-with-prompts/461931/10 "2023-11-03T20:20:37Z")
Understood, ChatGPT reused my use of MadMax and then it’s own prompt got denied. Bottom line this may get better or it may not ![DALL·E 2023-11-03 16.18.51 - Imagine a gritty cartoon character, a survivor of a post-apocalyptic world. His head is unusually tall, at least 50% taller than its wi](https://us1.discourse-cdn.com/openai1/original/3X/3/7/37926277ec980f6c399d17ee34bd6349df21580c.jpeg) ---
**Author:** ![MagnetMan](https://sea2.discourse-cdn.com/openai1/user_avatar/community.openai.com/magnetman/32/149130_2.png) [@MagnetMan](https://community.openai.com/u/MagnetMan)\ **Post date:** [November 4, 2023, 11:57pm UTC](https://community.openai.com/t/content-policy-issues-with-prompts/461931/11 "2023-11-04T23:57:56Z")
It clearly states on their website that all images created are free to be used, commercially without any limitations ---
**Author:** ![johncain194](https://sea2.discourse-cdn.com/openai1/user_avatar/community.openai.com/johncain194/32/147357_2.png) [@johncain194](https://community.openai.com/u/johncain194)\ **Post date:** [November 9, 2023, 2:31am UTC](https://community.openai.com/t/content-policy-issues-with-prompts/461931/12 "2023-11-09T02:31:53Z")
What I mean is not “free to use” or “commercially” without any limitations, but what I meant is you have many “limitations” imposed on you when you want to produce your images. ---
**Author:** ![Macha](https://sea2.discourse-cdn.com/openai1/user_avatar/community.openai.com/macha/32/171757_2.png) [@Macha](https://community.openai.com/u/Macha)\ **Post date:** [November 10, 2023, 7:57pm UTC](https://community.openai.com/t/content-policy-issues-with-prompts/461931/13 "2023-11-10T19:57:31Z")
Yep. Welcome to guardrails. You cannot use DALL-E 3 without limitations. The only image gen model without guardrails is stable diffusion. Also, I do want to point out AI imagery is free to use commercially, but US IP law will not copyright any AI generated content. Meaning anyone else can use the same image the you generated without consequence. ---
**Author:** ![Spongiform](https://sea2.discourse-cdn.com/openai1/user_avatar/community.openai.com/spongiform/32/246260_2.png) [@Spongiform](https://community.openai.com/u/Spongiform)\ **Post date:** [November 12, 2023, 11:41pm UTC](https://community.openai.com/t/content-policy-issues-with-prompts/461931/14 "2023-11-12T23:41:49Z")
I’ve been playing around with single word or short phrase prompts just to try to learn what dall-e associates with what. Two words that I couldn’t use were “fear” and “tentacle” Not together, just the singular word by itself. ---
**Author:** ![MagnetMan](https://sea2.discourse-cdn.com/openai1/user_avatar/community.openai.com/magnetman/32/149130_2.png) [@MagnetMan](https://community.openai.com/u/MagnetMan)\ **Post date:** [November 13, 2023, 12:12am UTC](https://community.openai.com/t/content-policy-issues-with-prompts/461931/15 "2023-11-13T00:12:08Z")
There’s a fascinating art to persuading the bots to create subject matter that is normally censored. Involves lying cheating and stealing, just kidding. If you can learn how to craft a prompt with descriptive words, that it favors, words that have the same meaning, as what you were looking for but within the boundaries of political correctness, non-discrimination and not offensive to the Wolk populace, you can perform miracles. I found this earlier today on a Bing related FB page, those are meatballs without sauce. ![IMG_1369](https://us1.discourse-cdn.com/openai1/original/3X/b/d/bd135fa65a5e40f39623038fd7b7cfd34f575b42.jpeg)