Error 715-123420 — fraud-detection block on all GPT-5 model deployments.

Sorin Buzatu 20 Reputation points
2026-06-22T20:09:36.4233333+00:00

We are encountering a new issue with deploying a GPT-5 or GPT-5.4 model in our subscription via Foundry!

I have tried to create an support ticket (complicated), without success even though we have a developer support plan subscription.

This is the issue which I need to be treated please with high priority:

All GPT-5-family model deployments fail instantly with error 715-123420

("An error occurred. Please reach out to support for additional assistance"),

while gpt-4o, gpt-4o-mini, and text-embedding-3-large deploy and run fine on

the SAME resource. Reproduced via Azure CLI (az cognitiveservices account

deployment create), so this is not a portal-only issue.

I have already ruled out every config-level cause:

  • Subscription a74ce9d7-8c96-4b79-9b6a-e696e6569f47 (Concentric Digital),

  PayAsYouGo, spending limit Off, state Enabled.

  • Resource: aIdocumentprocessor-openai, resource group TestEnvironment,

  region Poland Central.

  • Attempted: gpt-5.4 (version 2026-03-05), GlobalStandard, 150K TPM.
  • Quota IS available: OpenAI.GlobalStandard.gpt-5.4 limit 1000, used 0.
  • Region + resource both list gpt-5.4 GlobalStandard as available

  (az cognitiveservices account list-models confirms it).

  • Provider Microsoft.CognitiveServices is Registered.

This is the documented RTFP (Real-Time Fraud Protection) backend block.

Please escalate to the RTFP / Fraud Protection backend review team to remove

the fraud-signal block on GPT-5 / frontier-model deployments for this

subscription.

Azure AI Search
Azure AI Search

An Azure search service with built-in artificial intelligence capabilities that enrich information to help identify and explore relevant content at scale.


Answer accepted by question author
Anshika Varshney 15,625 Reputation points Microsoft External Staff Moderator
2026-06-25T19:56:28.7566667+00:00

Hello @Sorin Buzatu

We have received an update from the PG team that the issue has been rectified. Analysis and adjustments completed, please try again within 1-2hours.

Could you please check from your end and confirm whether the issue is now resolved?

If the updated response resolves your query, kindly consider clicking Upvote and Accept Answer so it can help others as well.

Looking forward to your response. Thankyou!

Thankyou!

Was this answer helpful?

1 person found this answer helpful.
0 comments No comments

1 additional answer

Sort by: Newest
  1. Sina Salam 31,456 Reputation points Volunteer Moderator
    2026-06-29T11:16:31.63+00:00

    Hello Sorin Buzatu,

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

    I understand that you are having Error 715-123420 — fraud-detection block on all GPT-5 model deployments.

    With careful evaluation of your scenario, the permanent resolution must come from Microsoft backend review because the deployment failure occurs after the you already validated the normal customer-controlled causes: subscription state, quota, model availability, deployment type, provider registration, and CLI reproduction. Since your checks pass and 715-123420 still blocks only GPT-5-family deployments, the issue is no longer fixable through portal, RBAC, quota, or resource recreation. - https://learn.microsoft.com/en-us/azure/foundry/openai/how-to/quota, https://learn.microsoft.com/en-us/azure/foundry/openai/quotas-limits, https://learn.microsoft.com/en-us/cli/azure/cognitiveservices/account/deployment?view=azure-cli-latest

    Because of the delay, if you're in production environment the action is to keep the workload running on an already approved model such as gpt-4o or gpt-4o-mini while escalating the waiting for Azure technical support for backend validation, risk, abuse-monitoring, or model-access review. You can also contact Priority Customer Support .

    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.

    Was this answer helpful?


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.