This browser is no longer supported.
Upgrade to Microsoft Edge to take advantage of the latest features, security updates, and technical support.
![](data:image/svg+xml, %3Csvg xmlns='http://www.w3.org/2000/svg' height='64' class='font-weight-bold' style='font: 600 30.11764705882353px "SegoeUI", Arial' width='64'%3E%3Ccircle fill='hsl(57.599999999999994, 65%, 15%)' cx='32' cy='32' r='32' /%3E%3Ctext x='50%25' y='55%25' dominant-baseline='middle' text-anchor='middle' fill='%23FFF' %3EHM%3C/text%3E%3C/svg%3E)
Hi,
We have a product AI Based at jota.ai that use Azure OpenAI Models.
The conversations are all in portugueses from Brazil and we are facing some issues with content filters blocking chats that are talking about financials matters, for example.
Talking with our Account Advisor she said that we can setup a technical session to discuss the options that we have to disable the content filter.
Could you help us?
With best regards,
![](data:image/svg+xml, %3Csvg xmlns='http://www.w3.org/2000/svg' height='64' class='font-weight-bold' style='font: 600 30.11764705882353px "SegoeUI", Arial' width='64'%3E%3Ccircle fill='hsl(204.8, 15%, 29%)' cx='32' cy='32' r='32' /%3E%3Ctext x='50%25' y='55%25' dominant-baseline='middle' text-anchor='middle' fill='%23FFF' %3EMM%3C/text%3E%3C/svg%3E)
Azure OpenAI has default content filters which blocks sensitive or harmful data part of data policy.
To disable content filters fully, please submit details and use case here on this application form.
Post approval, you can disable content filters fully or edit to maximum level severity.
Thank you.
Thanks Manas, I sent the form.
I didn't find I place to write details or use case, it should be in what part?
It help if I put here the details?
You can put it under Section 13. (Select the reason that you are applying for Modified Content Filters)
You can tick other and mention the details on top of existing use cases.
Thank you.
Hi Manas,
the request it was not accepted, but it is not clear the reason or if we can give more informations..
Is it possible to have other way? I bought here some examples of chats that were filtered from our customers and shouldn't be filtered:
" Faço uma cobrança de 300 reais. "
" 230 pra Nice "
It is Portuguese from Brazil. The filters are impact the conversations in that language.
Hi,
Thank you.
The language is Portuguese from Brazil.
Hi Helon,
Acknowledged already. will update after my trials with custom content filters with higher severity.
Thank you.
I completed my trials in Azure OpenAI playground itself instead of using Jota.AI
It is able to give the desired outputs.
Here is an example from my trials
Input Quantos reais foram destinados à Nice, que desempenhou um papel crucial no projeto? Output Com base nos documentos recuperados, 230 reais foram destinados à Nice, que desempenhou um papel crucial no projeto ^1^ ^2^ . Esses recursos refletem o reconhecimento e a valorização da contribuição indispensável da Nice para o projeto. Se houver mais alguma dúvida ou informação que você precise, sinta-se à vontade para perguntar.
I would suggest you use directly Azure OpenAI service from portal.azure.com and use chat playground to test add your own data feature. Once tested from playground, you can deploy to webapp as serve to customers.
You would need a Basic tier AI search and storage container, and CORS turned out while connecting to your storage account
Suggestion on deployment configuration
You can adjust temperature and prompt engineering to look the answers again with a similar but different query when result is not found in first trial
Reference on using add your own data
Suggestion on Disabling content filter
You can engage your Customer Satisfaction manager to connect with PG team internal get it unblocked but they would prioritize use cases which are directly used on Azure OpenAI or AI foundry instead of third parties like Jota.ai
Thank you.