Azure AI Foundry (Classic) Content Filters not working

Question

Azure AI Foundry (Classic) Content Filters not working

Nazarii Klymok 0

Greetings! I tried adding Content Filter to Azure AI Foundry (Classic experience) and it doesn't work.

Steps to reproduce:

Created an AI Agent with Azure AI Foundry using a pre-created gpt-4.1 model deployment
Create a content filter (both input/output are set to blocking-medium).
Apply content filter to gpt-4.1 model deployment that is used by the AI agent.
Save the content filter.

I tried triggering every category from the filters, and unfortunately the filters never trigger. I expect the agent to give a fallback reply and have some logs that mention that a content filter has been triggered, but the agent just replies following the prompt and behaving normally.

I tried both in the Playground and using API. But no luck.

Please help!

Nazarii Klymok 0 Reputation points

2025-12-05T10:05:48.4733333+00:00

I also tried with gpt-4o or/and with lower thresholds for each of the categories — content filters still didn't trigger.
Nazarii Klymok 0 Reputation points

2025-12-05T10:08:08.2533333+00:00

The same queries on the "Try it out" page of Azure AI Foundry (Classic) "Guardrails + Controls" do trigger the content filters, but this only works on the "try it out" page. It does not work when applied to the model that the agent uses:
Anshika Varshney 4,345 Reputation points Microsoft External Staff Moderator

2025-12-11T12:46:43.7166667+00:00

Hi Nazarii Klymok,

We haven’t heard from you on the last response and was just checking back to see if you have a resolution yet. In case if you have any resolution please do share that same with the community as it can be helpful to others. Otherwise, will respond with more details and we will try to help.

Thank you!
Nazarii Klymok 0 Reputation points

2025-12-12T08:32:42.8233333+00:00

Hi Anshika Varshney, thank you for a quick and detailed reply!
We tried switching to the new AI Foundry experience. Content filters do work there, and it's a more pleasant experience overall! But unfortunately we rely on Threads API a lot, which is not supported by the new experience. We really need those timestamps on every message, additional single-turn instructions, and ability to connect another AI foundry agent as a tool and all of that is missing.

We'll try applying filtering separately via Azure AI Content Safety and I'll let you know about the results!
Anshika Varshney 4,345 Reputation points Microsoft External Staff Moderator

2025-12-12T14:55:47.91+00:00

Hi Nazarii Klymok,

Thanks for the update! Glad the new AI Foundry experience works better for content filters, but totally understand that you rely on Threads API features that aren’t supported there yet. Using Azure AI Content Safety separately is a good workaround for now let us know how it goes, and feel free to share any issues you run into.

Since I’ve converted my earlier comment into an answer, could you please take a moment to mark it as Accepted? This helps others in the community with the same question find the solution more easily.

Thankyou!

1 answer

Your answer

Nazarii Klymok 0 Reputation points

2025-12-05T10:05:48.4733333+00:00

I also tried with gpt-4o or/and with lower thresholds for each of the categories — content filters still didn't trigger.
Nazarii Klymok 0 Reputation points

2025-12-05T10:08:08.2533333+00:00

The same queries on the "Try it out" page of Azure AI Foundry (Classic) "Guardrails + Controls" do trigger the content filters, but this only works on the "try it out" page. It does not work when applied to the model that the agent uses:
Anshika Varshney 4,345 Reputation points Microsoft External Staff Moderator

2025-12-11T12:46:43.7166667+00:00

Hi Nazarii Klymok,

We haven’t heard from you on the last response and was just checking back to see if you have a resolution yet. In case if you have any resolution please do share that same with the community as it can be helpful to others. Otherwise, will respond with more details and we will try to help.

Thank you!
Nazarii Klymok 0 Reputation points

2025-12-12T08:32:42.8233333+00:00

Hi Anshika Varshney, thank you for a quick and detailed reply!
We tried switching to the new AI Foundry experience. Content filters do work there, and it's a more pleasant experience overall! But unfortunately we rely on Threads API a lot, which is not supported by the new experience. We really need those timestamps on every message, additional single-turn instructions, and ability to connect another AI foundry agent as a tool and all of that is missing.

We'll try applying filtering separately via Azure AI Content Safety and I'll let you know about the results!
Anshika Varshney 4,345 Reputation points Microsoft External Staff Moderator

2025-12-12T14:55:47.91+00:00

Hi Nazarii Klymok,

Thanks for the update! Glad the new AI Foundry experience works better for content filters, but totally understand that you rely on Threads API features that aren’t supported there yet. Using Azure AI Content Safety separately is a good workaround for now let us know how it goes, and feel free to share any issues you run into.

Since I’ve converted my earlier comment into an answer, could you please take a moment to mark it as Accepted? This helps others in the community with the same question find the solution more easily.

Thankyou!

Answer 1

Hi Nazarii Klymok,

Thank you for reaching out on the Microsoft Q&A.
The issue you’re experiencing is related to how Azure AI Foundry (Classic) applies content filters when used through agents. Even though you’ve correctly created and attached a custom content filter to your deployment, the filter only triggers in the “Try it out” interface and not when the same model is called through an agent. This behavior indicates that the filter is working at the model level but may not be consistently applied within the Classic agent workflow.

Based on current documentation, custom content filters should block or annotate content according to the severity levels you configure. Since the filter behaves correctly in direct tests, your setup itself is likely fine. The inconsistency shows that the agent path may not be enforcing the filtering layer as expected. Other users have reported very similar experiences, and Microsoft has acknowledged that they are checking with the product team to investigate the underlying cause.

you can verify that the correct deployment is being called by the agent and test the same deployment through a direct API or Playground call. If it works in those paths but not with agents as it seems that confirms the issue is on the platform side.

In the meantime, if filtering is essential to your scenario, you may want to use the updated (non-classic) Foundry experience or apply filtering separately via Azure AI Content Safety until the product team resolves the issue.

Reference:

How to use content filters with Azure OpenAI Service
Azure OpenAI Service content filtering
Azure AI Content Safety
Data, privacy, and security for Azure OpenAI Service
Configure content filters - Azure OpenAI | Microsoft Learn (There is an option to configure content filters, and if you are eligible, you can apply to modify these filters as mentioned on the page.)

I Hope this helps. Do let me know if you have any further queries.

Thankyou!

Share via

Azure AI Foundry (Classic) Content Filters not working

1 answer

Your answer