Azure OpenAI resource temporarily blocked with HTTP 403: request for manual review

Question

Hello,

I am requesting a manual review of an Azure OpenAI resource that appears to have been temporarily blocked by an automated enforcement system. Based on the workload and the behavior I observed, I believe this is likely a false positive.

Task

The resource is used for a non-production data-engineering pipeline that improves cooking-recipe instructions. The model receives ordinary culinary content and rewrites it into clearer, structured preparation steps. The workload does not involve harmful content, security testing, attempts to bypass safety systems, or user-generated conversations.

The test running when the problem appeared was a small, targeted validation containing exactly five recipe records. The recipes involved normal cooking tasks such as:

• Correcting shrimp cooking times
• Clarifying caramel preparation and heat instructions
• Preserving crab ingredient identity
• Clarifying the blending and serving of salmorejo
• Consolidating baccalà preparation steps and gradual oil incorporation

Error

The test used the gpt-5.1 deployment with five concurrent records. Each record was allowed one initial generation and, if necessary, one narrowly scoped correction. This was not a large-scale or open-ended workload. The same configuration was functioning successfully earlier that day. It subsequently began returning the following response for every request:

Forbidden: Your resource has been temporarily blocked because we detected behavior that may violate our content policy.

Details:

• First observed: July 29, 2026, at approximately 12:31 UTC
• The same configuration was working at approximately 10:14 UTC
• API operation: Responses API
• Affected deployments: gpt-5.1 and gpt-5-mini
• Both deployments show a status of Succeeded
• The same 403 occurs from my application and directly from the Microsoft Foundry playground
• No API keys were intentionally rotated before the problem began
• No model responses were produced and no tokens were consumed by the rejected requests

Suspected Cause

I believe this may be a false positive for the following reasons:

1. The workload consists exclusively of ordinary cooking recipes and recipe-editing instructions.
2. The same resource and configuration worked successfully only a few hours earlier.
3. Both model deployments are affected, regardless of the requested recipe.
4. A basic request made through the Foundry playground also receives the same resource-level 403.
5. The requests are rejected before model processing rather than receiving a prompt-specific content-filter response.

Request

Could a Microsoft moderator please initiate a private conversation with me and escalate this to the appropriate Azure OpenAI provisioning or Trust & Safety team for manual review?

I can provide necessary IDs, numbers and more necessary information privately. If this is not the correct way to submit this request, please let me know exactly where I should turn to. Please also let me know if there is a way for me to resolve this issue on my own, I did not find a way to do so.

Thank you very much for your help and support,

Best, Felix

Answer 1

Hello Felix K,

Welcome to the Microsoft Q&A and thank you for posting your questions here.

I understand that your Azure OpenAI resource temporarily blocked with HTTP 403: request for manual review.

This is a resource-level enforcement block, not a deployment, quota, API key, SDK, model-health, or normal content-filter issue. The key indicator is that the same 403 occurs across multiple deployments and also from Microsoft Foundry playground, while deployments remain Succeeded and no tokens are consumed.

What you can do to resolve is to:

Stop all traffic to the affected resource and open an Azure Support request from the affected subscription in your Azure portal, asking Microsoft team to escalate the case to the Azure OpenAI / Foundry Models Trust & Safety / abuse-monitoring enforcement team for manual review, unblock assessment, and root-cause classification. You can also, contact Priority Customer Service.

There is no need to rotate keys, redeploy the model, tune quota, or change content filters as the primary fix. Normal content filtering is prompt/completion-level and returns content-filter-specific outcomes, while this scenario is a blanket resource-level 403 block. - https://learn.microsoft.com/en-us/azure/ai-foundry/openai/concepts/content-filter

I hope this is helpful. Please! Do not hesitate to let me know if you have any other questions, steps or clarifications.

---

Please do not close the thread by upvoting and accepting the answer if any part of it is helpful.

Answer 2

I think 403 resource temporarily blocked cannot be cleared from the user side, and the action is to stop calls and send the resource name, region, subscription ID, the UTC failure time, and the x‑ms-client-request-id to Azure support, and with the Developer plan the moderator can start the private thread for you.

Answer 3

Hi @Felix K

The HTTP 403 response ("Your resource has been temporarily blocked because we detected unusual behavior") is typically an automated protection mechanism triggered by Azure OpenAI. Unfortunately, the block cannot be cleared from the client side.

To help identify the cause, I'd recommend checking the following:

• Review your application's request patterns for sudden spikes, high concurrency, or retry loops that may have been interpreted as unusual behavior.
• Verify that your application is respecting your deployment's rate limits (TPM/RPM) and is using exponential backoff for 429/5xx responses instead of aggressive retries.
• Review recent prompts and completions to ensure they comply with the Azure OpenAI Content Policy, as repeated policy-triggering requests may also result in a temporary block.

If the resource remains blocked after reducing traffic and verifying your workload, you'll need to contact Microsoft Support (or request assistance from a Microsoft moderator if you're using a support plan that doesn't allow ticket creation). Include:

• Resource name and region
• Subscription ID
• Timestamp(s) of the failures (UTC)
• x-ms-client-request-id or apim-request-id from the failed requests
• A brief description of your workload and recent changes

These details allow the product team to review the block and determine whether the resource can be unblocked. Similar cases on Microsoft Q&A have required backend investigation by the Azure OpenAI team.

Please "Accept the Answer" if this information helped you. This will help us and others in the community as well.