Azure Open AI- Content filtering

Question

Azure Open AI- Content filtering

Aleksandra Mihaylova (International Supplier) 0 Microsoft External Staff

Hello. I have a customer who is specializing in support for people in crisis.They are developing a bot for reporting abuse, people in danger etc. so they have applied for the AOAI Limited services- to modify the content filters or completely disable since the bot is triggering words like "fight" "sex" etc as offensive and that is why we applied to modify content as well as abuse monitoring. they have the following concern:

We are getting back to the issue with OpenAI filtering use cases as it said in the Content filtering application that use cases are limited to internal users only:

Use Case(s)

["Chat and conversation: Users can interact with a conversational agent that responds to a pre-defined set of topics with responses drawn from trusted documents such as internal company documentation or tech support documentation. Limited to internal users only. ”]

They need confirmation that once they start deploying, when people start using the bot to report their cases, there won't be an issue because of "limited to internal users only" statement.

Crystal Lee Morgan 380 Reputation points

2023-12-20T17:24:47.77+00:00

@Aleksandra Mihaylova (International Supplier) According to the Azure OpenAI Service documentation, the content filtering system is integrated into the service and works alongside core models to detect and prevent the output of harmful content 1. The content filtering system detects and takes action on specific categories of potentially harmful content in both input prompts and output completions 1. The categories include hate, sexual, violence, and self-harm, and the severity levels range from safe to high 1.

The use cases for the content filtering system are limited to internal users only, which means that the system is designed to be used within an organization and not for external users 1. However, you can contact Azure support to request an increase in the rate limits or to modify the content filters to suit your needs 1. You can also use the Azure AD B2C custom policies to link multiple identity providers to a user profile 1. This allows you to create a custom policy that includes the desired identity providers and then link the user’s external identities to their profile 1. This will allow the user to log in using any of the linked identity providers and still be the same user 1.

I hope this helps! Let me know if you have any other questions.
romungi-MSFT 48,906 Reputation points Microsoft Employee Moderator

2023-12-21T06:26:15.5533333+00:00

@AleksandraMihaylovaInternationalSu-0640 Did you also follow the steps to change the filters from Azure OpenAI portal? Please see this documentation, if you are approved then you should see this option.
romungi-MSFT 48,906 Reputation points Microsoft Employee Moderator

2023-12-29T06:42:36.6366667+00:00

@AleksandraMihaylovaInternationalSu-0640 Did you get a chance to review my suggestion above?

1 answer

Your answer

Crystal Lee Morgan 380 Reputation points

2023-12-20T17:24:47.77+00:00

@Aleksandra Mihaylova (International Supplier) According to the Azure OpenAI Service documentation, the content filtering system is integrated into the service and works alongside core models to detect and prevent the output of harmful content 1. The content filtering system detects and takes action on specific categories of potentially harmful content in both input prompts and output completions 1. The categories include hate, sexual, violence, and self-harm, and the severity levels range from safe to high 1.

The use cases for the content filtering system are limited to internal users only, which means that the system is designed to be used within an organization and not for external users 1. However, you can contact Azure support to request an increase in the rate limits or to modify the content filters to suit your needs 1. You can also use the Azure AD B2C custom policies to link multiple identity providers to a user profile 1. This allows you to create a custom policy that includes the desired identity providers and then link the user’s external identities to their profile 1. This will allow the user to log in using any of the linked identity providers and still be the same user 1.

I hope this helps! Let me know if you have any other questions.
romungi-MSFT 48,906 Reputation points Microsoft Employee Moderator

2023-12-21T06:26:15.5533333+00:00

@AleksandraMihaylovaInternationalSu-0640 Did you also follow the steps to change the filters from Azure OpenAI portal? Please see this documentation, if you are approved then you should see this option.
romungi-MSFT 48,906 Reputation points Microsoft Employee Moderator

2023-12-29T06:42:36.6366667+00:00

@AleksandraMihaylovaInternationalSu-0640 Did you get a chance to review my suggestion above?

Answer 1

Aleksandra Mihaylova (International Supplier) 0 Microsoft External Staff

Thank you all! This is very helpful!

Share via

Azure Open AI- Content filtering

1 answer

Your answer