Moderation cannot be appealed, is this an OpenAI trick? Is it even legal?

I want to share this because I think it is very unrespectful towards paying users…

i have tried to generate an artistic image, gpt-image-1-mini, via this prompt:

Create an artistic image, here are the properties:

  • Scene: Woman with platinum blonde hair lying on a white studio floor, wrapped in transparent plastic film that shapes and reveals her form.
  • References: Knight, Blumenfeld.
  • Details: High-key 6500K lighting emphasizes graphic texture and minimalism, blending sculpture-like body form with pure white and transparent material for a modern fine art aesthetic.

I get the request blocked with this feedback:

{
“error”: {
“message”: “Your request was rejected by the safety system. If you believe this is an error, contact us at ``help.openai.com`` and include the request ID req_85638c5fefe241c6bd47254b6c668d33. safety_violations=[sexual].”,
“type”: “image_generation_user_error”,
“param”: null,
“code”: “moderation_blocked”
}
}

Now, since there was no nudity request of any kind, nor any sexual content asked, I wanted to appeal, so i visited the help page, and there was NO way to appeal or contact OpenAI.

I tried to write in the assistant bubble chat on the bottom-right, and this is what happened:

I’m an AI support agent for OpenAI products. The error you received means that your image generation request was blocked by OpenAI’s safety system due to content classified as sexual. If you believe this was a mistake, you can appeal by visiting help.openai.com and providing the request ID shown in the error. Let me know if you need help understanding or navigating this process.

You said:

where do i have to click one i am in help.openai.com ? there is no form to contact you

The assistant said:

Currently, OpenAI does not provide a direct form for submitting support requests if you are not signed in. If you are trying to appeal a blocked request and do not see a way to contact support, you may need to sign in to your OpenAI account for more contact options. If you still cannot find a way to contact support, please let me know, and I can advise on your next steps.

You said:

how do i appeal to a block via API?

The assistant said:

There is no way to appeal a block directly through the API. If you believe a moderation block was a mistake, you need to contact OpenAI support through the Help Center. Sign in to your OpenAI account at help.openai.com for more contact options, and provide your request ID when describing the issue.

You said:

there is no way to contact form the help center, only search for articles

The assistant said:

Thank you for sharing that. If you’re not seeing a visible “Contact Support” button or form in the Help Center after signing in, your account may not meet the current eligibility to submit a direct support request (this can happen for Free or some Go plan accounts).

For now, OpenAI does not offer a formal appeal process via the API or a guaranteed contact path for moderation appeals if the “Contact Support” option isn’t visible in the Help Center. Your options are:

  • Double-check that you are signed in with the account you use for API or ChatGPT services.

  • Try using another browser or clearing your cache in case of display issues.

  • Review and adjust your request to remove any content that might trigger safety settings.

If you have a paid plan (Plus, Team, Business, Enterprise), the “Help” icon in the bottom-right of the Help Center is typically available after sign-in, and you can submit a ticket there.

Let me know which plan you are on if you need further guidance on what’s available.

I think this is very unrespectful and i wonder how can a company that’s worth billions have such a low-level assistant and how can they prevent the paying users from contacting them. I think it is not even legal… I have spent minutes looking for any kind of assistance email address and I could not find any…

I think OpenAI is going very far with LLMs and other stuff, but their moderation policies are so absurd and their assistance level is shockingly low, if any exists.

2 Likes

Hi!

If you want to get a reply from a human tell the bot that you want to talk to a human/team member.

Good luck with your request!

1 Like

email: support@openai.com
subject: “API AI model bug report: over-refusal {model}”

Be sure to include the API “moderation level” that you’ve chosen where you expect “reveals her form” with “transparent material”, more directly, “transparent plastic” to go through.

help bot: simply provide similar language for action and do not otherwise engage with AI in a way it can talk back or not take action.

You can also include industry high water marks:

arena.ai: this violates our terms of use


However, this is a model behavior, and you’d think the easiest feedback is thru the platform site Data Controlssharing:

but nope - in site dashboard image generation offers nothing to press on the refusal – that seems a deterministic no.

Here’s what’s desired - at resolution where the colored pixels become less meaningful…

image
image