Custom content filters and blocklists for GPT 4o in Azure OpenAI

Question

I have GPT 4o deployed in North Central US region in my Azure OpenAI resource. I have created a custom content filter with custom blocklists (using instructions from https://learn.microsoft.com/en-us/azure/ai-services/openai/how-to/use-blocklists) via the Azure OpenAI Studio. I am able to see content filtering work when using GPT 35 Turbo. But it is not working with GPT4o.

Can you please confirm if there are any limitations in using custom content filters/blocklists with different GPT models.

Answer 1

Mayank Sharma Apologies for the delayed response and thanks for your time and patience. I have further checked and did not find any limitations on using blocklist for GPT-4o model.

Could you share any specific error details which helps in finding the root cause?

Answer 2

I am facing the same issue. Deployed gpt-4o-mini model:

• When using the DefaultV2 content filter, the content filter works.
• When using custom content filter, the content filter does not work at all

Here is the bicep used to create the custom content filter

param location string = 'australiaeast'
param openAiServiceName string = 'openai-service'
param raiPolicyName string = 'content-filtering'
param blocklistName string = 'blocklist'

// Resources
resource openAiService 'Microsoft.CognitiveServices/accounts@2024-10-01' existing = {
  name: openAiServiceName
}

resource raiPolicy 'Microsoft.CognitiveServices/accounts/raiPolicies@2024-10-01' = {
  parent: openAiService
  name: raiPolicyName
  location: location
  properties: {
    basePolicyName: 'Microsoft.DefaultV2'
    contentFilters: [
      {
        name: 'Violence'
        blocking: true
        enabled: true
        severityThreshold: 'Low'
        source: 'Prompt'
      }

      {
        name: 'Violence'
        blocking: true
        enabled: true
        severityThreshold: 'Medium'
        source: 'Completion'
      }
      {
        name: 'Hate'
        blocking: true
        enabled: true
        severityThreshold: 'Low'
        source: 'Prompt'
      }
      {
        name: 'Hate'
        blocking: true
        enabled: true
        severityThreshold: 'Medium'
        source: 'Completion'
      }
      {
        name: 'Sexual'
        blocking: true
        enabled: true
        severityThreshold: 'Low'
        source: 'Prompt'
      }
      {
        name: 'Sexual'
        blocking: true
        enabled: true
        severityThreshold: 'Medium'
        source: 'Completion'
      }
      {
        name: 'Selfharm'
        blocking: true
        enabled: true
        severityThreshold: 'Low'
        source: 'Prompt'
      }
      {
        name: 'Selfharm'
        blocking: true
        enabled: true
        severityThreshold: 'Medium'
        source: 'Completion'
      }
      {
        name: 'Jailbreak'
        blocking: true
        enabled: true
        severityThreshold: 'Low'
        source: 'Prompt'
      }
      {
        name: 'Indirect Attack'
        blocking: true
        enabled: true
        severityThreshold: 'Low'
        source: 'Prompt'
      }
    ]
    customBlocklists: [
      {
        blocking: true
        blocklistName: blocklistName
        source: 'Prompt'
      }
    ]
    mode: 'Default'
  }
}