Azure OpenAI deployments blocked by 715-123420 since Oct 11, support case unassigned

Suchang Ko 0 Reputation points
2026-10-27T06:24:13.8466667+00:00

Since 2026-10-11, every model deployment on my Azure OpenAI resource (koreacentral, GlobalStandard) fails with:

715-123420 - Our system has detected this request as unusual activity for your account. If you are confident this is in error, please contact support.
  • Models tried: gpt-6-luna, gpt-6.1-sol, gpt-4.1-mini. It fails on the first deployment.
  • Quota is available (0 of 250), the resource is healthy, and the subscription is active.
  • Latest failure: 2026-10-26 02:41 UTC.
  • Support case 2610150060002155 has had no engineer for 12 days. An earlier case was closed as a quota request.

Could a moderator review this or route it to the team that handles these blocks? I can share the subscription ID and correlation IDs privately.

It is a personal dev/test subscription for a small tool that classifies my own Slack mentions.

Azure OpenAI in Foundry Models
0 comments

2 answers

Sort by: Newest
Most helpful Newest Oldest

  1. AI answer

    2026-10-28T04:40:06.37+00:00

    Error 715-123420 requires Microsoft Support to escalate the existing case for a backend security review; add a message explicitly requesting a collaboration/escalation to the Security or Azure AI Foundry service team for review and removal of the temporary deployment block.

    Update and escalate the existing case

    1. In the Azure portal, open Help + support → All support requests.
    2. Select case 2610150060002155.
    3. Select New message, and submit wording such as:

      All Azure OpenAI model deployments fail immediately with 715-123420, including gpt-4.1-mini, despite available quota and a healthy resource.
      Please reclassify this from a quota request and create a collaboration/escalation with the Security/Azure AI Foundry service team for backend review of the temporary deployment block.
      First occurrence: 2026-10-11. Latest occurrence: 2026-10-26 02:41 UTC. Region: Korea Central. SKU: GlobalStandard. This is blocking the first deployment on the resource.

      Similar resolved cases record Support creating a collaboration with the Security team, after which the backend block was remediated.^1^
    4. Attach the deployment failure details:
      • UTC timestamp of each representative failure
      • Correlation/request ID
      • Azure OpenAI resource name and full resource ID
      • Subscription ID
      • Region, deployment SKU, model, and model version
      • Screenshot or exported deployment error
      Send subscription and correlation details only through the private support case—not in the public Q&A thread. Azure’s support guidance recommends supplying the start time, reproduction steps, and relevant logs while excluding confidential information from public/problem-description fields.^2^
    5. Because the case remains unassigned, select Change severity on the support request and raise it to the highest level accurately reflecting the impact and allowed by the support plan. An unassigned request permits changing severity directly.^3^

    Avoid further deployment troubleshooting

    Do not delete and recreate the resource, repeatedly retry deployments, or continue switching models and regions. A closely matching 715-123420 case states that these actions do not clear the restriction and that the block must be reviewed and released by Microsoft.^4^

    Your gpt-4.1-mini failure also rules out the workaround reported in cases where only newer model families were affected. Available quota does not resolve this particular backend restriction.

    If the case still receives no assignment

    Create a second support request specifically about managing an existing support request, reference the original case number, and ask for assignment rather than opening another quota case. Microsoft documents this as the escalation path when assistance is needed with an existing request.^3^

    A Q&A moderator can flag or escalate a matching thread, as occurred in another 715-123420 case, but cannot directly remove an account/subscription-level backend restriction.^4^ Keep the support case open until deployment succeeds, then perform one manual portal deployment before returning to any automation; that sequence was recommended after another subscription was unblocked.^5^

    References

    1. 715-123420: An error occurred. Please reach out to support for additional assistance. - Microsoft Q&A
    2. Create an Azure support request
    3. Manage an Azure support request
    4. Model Deployment error: 715-123420 - Microsoft Q&A
    5. Model deployment blocked with error 715-123420 "unusual activity" - Microsoft Q&A
    AI-generated content may be incorrect. Read our transparency notes for more information.

    Was this answer helpful?

  2. Jerald Felix 18,840 Reputation points Volunteer Moderator
    2026-10-28T02:16:33.9166667+00:00

    Hello Suchang Ko,

    Greetings! Thanks for raising this question in the Q&A forum.

    Error 715-123420 ("Our system has detected this request as unusual activity for your account") is not a quota, permission or configuration problem. It comes from Azure's automated risk and fraud protection checks, which run when a model deployment is validated, before anything is provisioned. That matches what you see: the same error on every model (gpt-6-luna, gpt-6.1-sol, gpt-4.1-mini), quota showing 0 of 250 used, a healthy resource and an active subscription. When this check flags a subscription, the block applies to all deployments on it, and it cannot be cleared from the portal, by changing regions, or by redeploying. Only the Microsoft team that owns the risk review can lift it. This is the same pattern other users have reported with this code, so your analysis is correct. I cannot see your subscription or lift the block from this forum, so the goal is to get your case in front of the right team.

    Here is what I suggest:

    Stop retrying and do not delete or recreate resources. Repeated failed deployments, or deleting and recreating the resource, can make the activity pattern look worse to the automated system. Keep the current resource as it is.

    Get your support case routed to the right queue. Your case 2610150060002155 has no engineer assigned, and your earlier case was closed as a quota request, which is likely why it went to the wrong team. Open a new support request (or update the existing one) with these exact settings: Issue type Technical, Service Azure OpenAI in Foundry Models (or Azure AI Foundry), and a problem type about model deployment failures. Do not choose a quota request type.

    Use wording that points at the risk review. In the description, write that deployments fail with 715-123420, that you suspect an automated fraud or risk protection (RTFP) hold on the subscription, and that you ask for it to be reviewed and lifted. Include the subscription ID, the resource name and region (koreacentral, GlobalStandard), the date it started (2026-10-11), the latest failure time (2026-10-26 02:41 UTC), the correlation IDs from the failed deployments, the models you tried, and your existing case number 2610150060002155.

    Explain your use case briefly. Add one or two lines that this is a personal dev and test subscription for a small tool that classifies your own Slack mentions, with low volume. Reviewers use this to confirm that the block was applied in error.

    Check that the subscription itself is in good standing. In the portal, look at Cost Management + Billing for any payment issue, spending limit or billing hold, and at Azure Policy for any Deny assignment on Foundry or Cognitive Services deployments. These are unlikely to be the cause given the error, but ruling them out saves time in the case.

    Escalate the stalled case. In the portal, open the support request and use the option to add a note asking for escalation, because 12 days without an engineer is outside normal response times. If your subscription has a support plan, you can also set the severity accordingly. If you have no paid support plan, you can still open requests for billing and subscription issues at no cost, and you can ask in the case for the issue to be handled as a subscription management or fraud review request. You can also ask a Microsoft moderator in this thread to route it, but account level actions still go through support.

    Use a workaround while you wait, if you can. If you have another subscription (for example a new pay-as-you-go one in good standing), you can test deployments there. Note that creating many new subscriptions to get around the block can itself trigger more flags, so use only one and keep the usage normal.

    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.