OpenAI Azure content filtering

Question

Hello,

Can someone provide some guidance on whether the content filter for Azure OpenAI can be removed?

We're finding it somewhat sensitive with what and when it chooses to moderate a response back and thus causing a 400 server status response.

See this example where I simply spell out my name letter by letter and the response I get back is a 400 status mentioning that the response was filtered due to the prompt triggering Azure OpenAI's content management policy. screenshot example

Any ideas on how to remove the content filter all together or explanation for why this bug is occurring and what can be done to make the filter as least strict as possible. This seems to happy consistently when spelling out my name ' A n d y' letter by letter as fed into the prompt as text.

Any help would be great.

thanks,

Answer 1

Hi Jeremy Lau

Please modify the prompts slightly or moderately if they are flagged as harmful content. Here in this example - you can simply use "ANDY" instead o "A n D y". You can find all harmful and non-harmful scenarios here. content filtering. To disable content filtering or use advanced content filtering, you need to submit details on advanced content filtering form and validate your use case.

Thank You.

Answer 2

Hi @Saideep Anchuri

The issue here though is that the input A N D Y is derived from the voice transcription when a caller spells out their name letter by letter which is a legit use case.

So therefore we're wondering why it is the case that it's throwing a content filtering 400 status which seemingly seems like a normal input. Is this a bug or is there something that can be done from the filtering side of things.

thanks,

Jeremy