How can we turn off content moderation policy for LLMS deployed as endpoints in AI Azure foundry?

Hi Amogha Subramanya,

You cannot fully turn off Azure AI Foundry’s content moderation for LLM endpoints by default, as Microsoft enforces baseline filters for safety. While you can create custom content filter configurations and lower the severity levels, these still apply “Annotate and Block” for certain categories. To switch to “Annotate Only” or disable filtering for specific harm categories, your organization must request an exception by applying through the Azure OpenAI Limited Access Review: Modified Content Filters form. Approval for this capability is typically granted only to managed enterprise customers, and Microsoft generally requires engagement through your Azure account representative. In your request, you should include context about your pharmaceutical use case and examples of false positives to justify why you need “Annotate Only” mode. Once approved, you will be able to configure content filters at a granular level and set them to “Annotate Only,” or even disable blocking entirely for certain categories, either through the Azure AI Foundry portal or by applying custom filters at runtime using the x‑policy‑id header.
For more information: Limited access for Azure OpenAI Service
Limited Access features for Azure AI services