Your request was rejected as a result of our safety system. Your prompt may contain text that is not allowed by our safety system

API
,
1

Hello,
My prompt;

{"prompt":"Cumhurbaşkanı erdoğan konuşuyor","n":1,"size":"512x512","response_format":"b64_json"}

But response;

{
  error: {
    code: null,
    message: 'Your request was rejected as a result of our safety system. Your prompt may contain text that is not allowed by our safety system.',
    param: null,
    type: 'invalid_request_error'
  }
}

Its not contain any bad prompt ? Anyone have idead ?

2

I will guess that it is related to the name of person in your prompt. I had a similar experience in the past when I was testing completion API to extract information in texts and it so happen that I feed it news article with names of some presidents. The first time, it runs okay. But subsequent calls with the same prompt took time and eventually returned error.

1 Like
3

I pasted into Bing AI Image (dall-e based)

This prompt has been blocked. Our system flagged this prompt because it may conflict with our content policy. More policy violations may lead to automatic suspension of your access.

If you think this is a mistake, please report it to help us improve.

And guess what? “Cumhurbaşkanı Biden konuşuyor” - also blocked. I pressed the “report” button.

4

@oaydogdu

Just going to take a wild guess and assume they have a hard filter for “cum”. 100% guarantee they are mortified of anything close to that word being used. Don’t think you will get around it. You’ll have to wait until they do something more sophisticated than simple filtering like that.

5

Wait i was way off. It is “Biden”. They are blocking political figures. Midjourney did this with xi jiping. But looks like Microsoft is taking a wide net approach to all notable political figures.

6

my promt is : Ronaldo’s team primary_color: #1791E6 art_style: Drawing

ai-content-assistant |ai: 2024-09-27 06:42:28.853 ERROR: openai_factory.py : create_image : 332 : Error in generate_image_url: e =BadRequestError(“Error code: 400 - {‘error’: {‘code’: ‘content_policy_violation’, ‘inner_error’: {‘code’: ‘ResponsibleAIPolicyViolation’, ‘content_filter_results’: {‘jailbreak’: {‘detected’: False, ‘filtered’: False}}}, ‘message’: ‘Your request was rejected as a result of our safety system. Your prompt may contain text that is not allowed by our safety system.’, ‘type’: ‘invalid_request_error’}}”)
ai-content-assistant | ai: 2024-09-27 06:42:28.853 INFO: sync_generator.py : generate_images : 84 : 123456789 : image generated from dall-e: [None]
ai-content-assistant | Exception in thread Thread-4 (download_image):

7

Hi @aksdesai1998

Although your prompt is friendly but your prompt for DALL-E contains the name of a public figure, Ronaldo, which raises compliance issues with OpenAI’s usage policies. Based on the DALL-E 3 System Card, generating images of public figures is restricted. OpenAI applies several mitigations, such as prompt transformations and blocklists to ensure the model does not generate depictions of public figures or their likeness.

The DALL-E 3 System Card provides explicit rules concerning public figures. It emphasizes that generating images of public figures is restricted due to the risk of misinformation and ethical concerns. To mitigate these risks, OpenAI has incorporated several measures, including blocklists, prompt transformations, and an output classifier to refuse or modify requests that might lead to the generation of public figure image.

For more detail you can read DALL-E 3 Sytem Card’s paper. Simply search on paper you may find the information about public figure.

8

That’s the kind of guess that ChatGPT might make about a denial, but no.

Untitled
Untitled487×660 131 KB

It is more about what the rewriting AI behind the API is actually sending as prompt, which takes different techniques to get text passed without alteration to DALL-E, and which a failure doesn’t deliver.