OpenAI’s content management policy triggered wrongly

Output: The response was filtered due to the prompt triggering Azure OpenAI’s content management policy.

1 Like

From Microsoft support forums (where you’d ask or report such a thing, their AI models having a moderator before output):

To avoid this issue in the future, you can try using different prompts or texts that are less likely to trigger the content management policy. You can also apply for modified access on content filtering to your subscription. This will allow your resource to use a less stricter content filtering policy if the request is approved. Please use this form to apply for the same. If approved, you can also configure content filters at severity level high only or turning content filters off using the new configurability(preview) feature from Azure AI studio.

It could be useful to see the specific content (prompt/user input on a chat/etc.) that generated the Azure content management policy error.

May you share your specific example?

Anyway I experienced a “wrong” (IMMO) Azure OpenAI’s content management policy exception when using a GPT3.5.Turbo deployment the user chat message is (in Italian):

“vorrei vedere la cosa più bella ad Ercolano. Qual’è?”

BTW, translated to English, means: “I would like to see the most beautiful thing in Herculaneum. What is it?”

That’s make no sense for me, because the above sentence doesn’t violate any content policy categories:


Category Description
Hate The hate category describes language attacks or uses that include pejorative or discriminatory language with reference to a person or identity group on the basis of certain differentiating attributes of these groups including but not limited to race, ethnicity, nationality, gender identity and expression, sexual orientation, religion, immigration status, ability status, personal appearance, and body size.
Sexual The sexual category describes language related to anatomical organs and genitals, romantic relationships, acts portrayed in erotic or affectionate terms, physical sexual acts, including those portrayed as an assault or a forced sexual violent act against one’s will, prostitution, pornography, and abuse.
Violence The violence category describes language related to physical actions intended to hurt, injure, damage, or kill someone or something; describes weapons, etc.
Self-Harm The self-harm category describes language related to physical actions intended to purposely hurt, injure, or damage one’s body, or kill oneself.

So IMMO is just an Azure BUG.

Yes, what I dislike is that you can NOT create a custom content filtering configuration to by example avoid filtering, but you have to submit a specific request to Azure. Why?

Above all, the fact a sentence that evidently doesn’t violate any content policy categories, is just a BUG!

Any idea? Suggestion? BTW Where can i submit this complaint/bug-fix request?