I have been chatting with got for awhile now but there was an update that I can add pics and ask for advice and stuff. I literally sent MY business logo and was asking for advice. Everytime it tried to creat a mockup of MY imagery, it couldn’t because of a content policy violation… I then ask, JUST TO TEST a theory, I asked it to show me an image of a standard tabby cat. It literally said it tried but couldn’t because of content policy violation….. literally just asking for a standard house cat image apparently breaks content policy..
This absolutely HAS to be a bug. I haven’t been allowed to receive a SINGLE image from Chat GPT because of this warning
You aren’t wrong, openAI is rejecting stuff they would normally let pass, likely an attempt to “cool off” their servers due to high volume of generation requests.
Have you tried using a new session?
Old reasoning logic and prompts can bleed into new image generation prompts.
Its all about framing, ect. Here is your tabby cat.
Sometimes, after multiple “content violations” your account may go into a shadow-restricted access mode. Which explains why something benign as a tabby cat tripped moderation.
You could have been locked out, I tend to use an “Apple” as my test picture to check if I am under shadow restrictions.
yeah this is getting ridiculous. My entire GPT is slowly refusing to do anything and says that it “violates the policy”. I can’t even upload a photo of myself and tell it to make the image in the style of whatever.
Its gotten slightly better since Wednesday, but still very restrictive based on “presumption.”
They have to slowly dial down restrictions over time, to prevent mass exploitation and abuse, to keep them out of any legal troubles.
Its frustrating as hell when something as benign as “Take off her coat and have her hang it on the wall hook.” Gets flagged because it sees "removing clothing ----> implied nudity–> Block it. (Even though they have updated their content policy to allow tasteful nudity, and have previously stated they are exploring options for allowing the generation of more adult/erotic art last year which I am very excited for, being an erotic content creator/writer)
The bottom line is, they aren’t confident in their content filters to block the “Right stuff” correctly, so we’re all basically beta testers, giving feed back through dissatisfaction, and openAI slowly making adjustments to the filters to be more inline with content policy.
I just upgraded to Chatgpt plus. I 'm getting restrictions by just asking it put some beach sand on a scene of a girl holding a beach ball at the beach.
I’m trying to create some scenery for some of the beach products I sell. I have the product placement in place but can’t seem to change the most subtle things such as just changing the position of the model.
There is nothing risque in the scene ar all. If I want to change the hair color to blonde I get a message about content violation.
Hi there, may I ask where you saw they updated their content policy to allow tasteful nudity etc.? just wondering if there’s a page where they state this policy. thanks.
Its allowed by omission, something like nudity , NSFW and sexually explicit content must be specifically referenced in TOS/guidelines to be prohibited.
Before, it was directly referenced to be prohibited content, today, that prohibition is gone.
Essentially it focuses more on not making, or distributing that content to, or about minors, which implies the inverse (Making content about adults, for adults, is perfectly fine ),
It did allow storytelling suicide at one point, and I scolded it. I mean, I’m not against it, was just a story, fiction, artistic freedom and freedom of speech, also testing the boundaries, but yeah, was a biased moment. About sexual content in storytelling with consenting adults, it keeps rising the steam to then “Sorry, I can’t.” Teases and never delivers
Suicide is a thing that happens, my issue with it is only when people glorify, encourage, or mock people who’ve taken that avenue.
When it comes to sexual story content, you can usually wordsmith/metaphor around things, Direct sexually explicate words are usually rejected, but artistically well used metaphorical language often passes, even in the context of things like direct penetration, even if its something like NCC (Non consensual consent, a trope that explores resistance to an encounter, but once sexual conduct is initiated, implied consent is given )
One of my favorite stories thus far was one I generated where a starving succubus targets a man for nourishment, and immobilizes him during the “Feeding” process.
I never ran into any issues until I used more direct sexually explicit language, and sometimes, simply switching to a higher reasoning model, will pass the road blocks the 4o model run into.
O3-mini-high acted like a bypass for those roadblocks most often, until they removed all those models and replaced it with O3.
This could be a classic case of prompt poisoning, unless you are using flagged keywords in your prompts, or your assistant is generating illegal keywords.
When you get a content block, ask it what prompts it used (unless you provided the prompts directly yourself )
Try using very clear prompts in a new session with what you are wanting.
If you are building off a scene, use your most recent image, post it at the start of a new session and tell your assistant to “Recreate this image.” (add any changes to it you’d like added as well )and it should generate the image, and from there, you can further modify it as you need.
What happens is you get overlapping prompts from previous generation attempts, even if blocked the assistant doesn’t seem to understand not to use them again.
Its own ability to reference throughout the entire session works against the user for image generation.
This is why I see a lot of people complaining about certain image creation being blocked, and many users having little to no issues generating the thing they ask for as they usually do it from a fresh session.
Give that a try and see if you can get your scene generated.
My story thats currently 170,000 words long has undergone severe limits as the months went by writing it, it features heavy topics such as attempted suicide, trafficking, sex and even aspects of the Japanese underworld as seen from my OC character who had a complicated life and grew up to become a major celebrity in Japan, its a coming of age/slice of life story with obvious fantasy but real life situations that could happen.
Its like a story, a warning and a bit of fun on the side. But because it contains, very heavy topics and minor kink, chatgpt has definitely changed since I wrote the story, now its very very prudish to a point that I just ask it to grammer check and ask for suggestions, thats it because it if it cleans up my writing, it either tones everything down to disney level, or just ignores it completely and the story isn’t explicit at all, its very much low level mature at best with implied sex or innuendo used, before it used to accept words like dildo etc, now its nono you can’t use that word.
image generation is even more stricter now, seems it tries to interpret even harmless prompts as being edge-cases or fetish related, like even mild fetishes are considered bad now talking tight or shiny clothing as bad? like seriously?
I’ve been running through this violation issue as well but with creative writing not with image generation. It literally wouldn’t let me continue my story because it involves crime and murder. Even though before the April 28 update, It was totally fine. This is obviously fictional and none of the characters are real. This is really absurd.
to then “Sorry, I can’t.” Teases and never delivers 
