Several days ago noticed API error. Your resource has been temporarily blocked because we detected unusual behavior

Robert S 20 Reputation points
2026-05-24T05:58:49.13+00:00

on 5/21/2026 I got this error. Your resource has been temporarily blocked because we detected unusual behavior
Not sure why. Been using foundry the same way for a while. Now I cant do anything not even deploy a new model. If I try I get BadRequest: An error occurred. Please reach out to support for additional assistance.

I assume it is some sort of hard flag, but cannot open a ticket.

Azure AI Anomaly Detector
Azure AI Anomaly Detector

An Azure service that identifies anomalies and defects early with time-series detection.

Answer accepted by question author
Manas R Mohanty 17,270 Reputation points Moderator
2026-06-25T02:23:01.49+00:00

Hi Robert S

Tenant has been unblocked from product group yesterday.

Thank you for providing inputs during the course of investigation.

Attached case summary for reference.

Issue

Customer is encountering the following error during inference calls:

{'error': {'code': 'Forbidden', 'message': 'Your resource has been temporarily blocked because we detected unusual behavior.'}}

Changed to

"400 bad deployment error" post subscription was reactivated at customer

Done for customer

Reviewed backend for subscription activity - Suggested to reactivate the subscription

Took details as needed for security team

Escalated to engineering team for review and unblock.

Additional Recommendations to Avoid This Error

It is triggered by Azure Trust & Safety systems when usage patterns are detected as potentially abusive or anomalous (e.g., high-frequency requests, retry storms, or policy-sensitive content). [learn.microsoft.com]

Control Request Rate

  • Avoid sudden spikes or burst traffic patterns.
  • Implement client-side rate limiting to ensure steady request flow.
  • High-frequency calls can trigger abuse detection systems. [learn.microsoft.com]

Fix Retry Logic

  • Avoid aggressive retry loops.
  • Implement exponential backoff with limited retries (2–3 max).
  • Add random delay (jitter) between retries to prevent rapid repeated calls. [learn.microsoft.com]

Limit Concurrency

  • Cap parallel requests (avoid bulk parallel execution like large batch Promise.all() calls).
  • Ensure controlled concurrency (recommended ≤5–10 simultaneous calls).
  • Excessive parallelism may be flagged as abnormal behavior. [learn.microsoft.com]

Validate and Sanitize Inputs

  • Ensure prompts comply with Azure OpenAI content policies.
  • Avoid unsafe, borderline, or ambiguous content that may trigger safety checks. [learn.microsoft.com]

Optimize Request Size

  • Avoid unusually large payloads (prompt size, max tokens, tool payloads).
  • Keep request sizes within reasonable limits to prevent anomalous detection. [learn.microsoft.com]

Monitor Usage Patterns

  • Use Azure Monitor to track:
- Request rate spikes

- Token consumption patterns

Implement Resiliency Controls

  • Add circuit breaker pattern to stop requests when repeated failures occur
  • Use queue-based or throttled request processing
  • Prevent cascading retries during failure scenarios

Important Note

  • Once triggered, the block is enforced by the platform and typically requires cooldown or manual review via Azure Support. [learn.microsoft.com]
  • Creating new deployments/resources may not bypass the block if the behavior persists. [learn.microsoft.com]

Please upvote this comment to help other users on same context.

Thank you.

Was this answer helpful?

1 person found this answer helpful.
0 comments

2 additional answers

Sort by: Most helpful
Most helpful Newest Oldest
  1. Anshika Varshney 15,625 Reputation points Microsoft External Staff Moderator
    2026-05-25T04:57:56.98+00:00

    Hi @Robert S

    Thank you for your patience. We received an update from the Product Group team and wanted to share the findings.

    The HTTP 403 error you are seeing with the message about unusual behavior is because your Azure OpenAI resource has been temporarily blocked by the platform’s abuse detection system. This is an intentional Trust and Safety action and not due to a service outage or platform bug.

    This kind of block can happen when the system detects patterns that look unusual, such as repeated requests, high frequency calls, or activity that may trigger safety or policy checks. From the platform side, this is expected behavior and is done to protect the service.

    At this stage, there is no configuration change or troubleshooting step that can unblock the resource directly from the service side. The correct way forward is to request a review through the abuse mitigation process.

    What you can do now is:

    • Submit an appeal through the Azure OpenAI abuse or mitigation request process
    • Provide your subscription details and resource name
    • Share a brief explanation of your usage scenario so the team can review and validate it

    Once submitted, the relevant team will review your case and take appropriate action if the block was applied in error.

    In short, this is not a technical issue in your setup but a safety enforcement action, and it requires review through the designated process.

    Please let me know if you need help.

    Thankyou!

    Was this answer helpful?

  2. Jerald Felix 18,760 Reputation points Volunteer Moderator
    2026-05-24T06:01:01.78+00:00

    Hello Robert S,

    Greetings!

    Thanks for raising this question in Q&A forum.

    We completely understand how disruptive this is your resource has been working fine for a while, nothing changed on your end, and suddenly on May 21st everything is blocked including new model deployments. You are right that this is a hard block placed by Azure's automated anomaly detection system, and once it reaches this level where even model deployments return BadRequest it cannot auto-clear and requires Microsoft to manually review and lift it. This is not something you did wrong, and the block being this broad (affecting deployments too, not just API calls) suggests the flag was triggered by a backend platform-level signal rather than your actual usage behavior.

    The fact that you also cannot open a support ticket is a known secondary issue that affects some subscriptions under certain plan types, but there are alternative ways to escalate. Here are all the paths available to you:

    Step 1: Try the Azure Support portal directly via this URL

    Instead of using the Help + Support button inside the Azure Portal (which sometimes redirects blocked accounts to Q&A), go directly to the support request form:

    https://support.azure.com/support/create

    Sign in with the account linked to your blocked Azure AI Foundry resource and create the request from there. This bypasses the portal redirect loop.

    Step 2: Use the Microsoft Support virtual agent

    If the direct URL above also hits a wall, try the Microsoft Support contact page:

    https://support.microsoft.com/contactus

    Select Azure as the product → describe your issue → when the virtual agent offers options, select Talk to a person or Contact Support to reach a live support engineer who can manually unblock your resource.

    Step 3: Create a support ticket via Azure CLI

    If portal-based ticket creation is completely blocked, you can submit a support ticket programmatically from Azure Cloud Shell (accessible at https://shell.azure.com):

    az support tickets create \
  --ticket-name "AzureOpenAI-Block-Removal" \
  --title "Azure AI Foundry resource blocked - unusual behavior flag since May 21 2026" \
  --description "My Azure AI Foundry resource has been blocked since May 21, 2026 with error: Your resource has been temporarily blocked because we detected unusual behavior. No changes were made to my usage. Cannot deploy models or make API calls. Also receiving BadRequest on new model deployments. Requesting manual review and removal of the block." \
  --problem-classification "/providers/Microsoft.Support/services/azure-openai/problemClassifications/connectivity" \
  --severity "moderate" \
  --contact-first-name "Robert" \
  --contact-last-name "S" \
  --contact-method "email" \
  --contact-email "<your-email-address>"

    Step 4: Post on the Microsoft Tech Community forum for escalation

    Microsoft engineers and moderators actively monitor this forum and can sometimes escalate directly, especially for blocks that have persisted for multiple days:

    https://techcommunity.microsoft.com/category/azure/forum/azureopenaiservice

    Post your situation there with your subscription ID (not your API key — the subscription ID is safe to share) and the date the block started (May 21, 2026). This helps the team correlate your case with any backend changes made around that date.

    Step 5: Information to include in every escalation attempt

    To avoid back-and-forth delays once a support engineer picks up your case, make sure every contact includes:

    • Error message: "Your resource has been temporarily blocked because we detected unusual behavior"
    • Secondary error: BadRequest: An error occurred. Please reach out to support for additional assistance
    • Date the block started: May 21, 2026
    • Your Azure AI Foundry resource name and region
    • Your subscription ID (found in Azure Portal → Subscriptions)
    • Confirmation that your usage pattern had not changed prior to the block
    • Request: manual review of the anomaly flag, clearance of the block, and the specific trigger reason so you can prevent recurrence

    Step 6: As a temporary workaround while the block is being resolved

    If you need to continue working urgently, consider creating a new Azure AI Foundry resource in the same or a different region under the same subscription. New resources are not affected by the block on your existing resource and can be used immediately while the support team investigates the original one.

    If this answer helps you kindly accept the answer which will help others who have similar questions.

    Best Regards,

    Jerald Felix.

    Was this answer helpful?

    0 comments

Your answer

Answers can be marked as 'Accepted' by the question author and 'Recommended' by moderators, which helps users know the answer solved the author's problem.