Freigeben über

Azure AI Foundry model deployments blocked – error 715‑123420

Nico Schröder 1 Zuverlässigkeitspunkt
2026-04-20T08:04:00.7933333+00:00

We are unable to deploy any model in Azure AI Foundry. All deployment attempts fail immediately with:

715-123420: An error occurred. Please reach out to support for additional assistance.

Impact

  • Affects all models (e.g. Mistral Large 3, OpenAI, others)
  • Reproduces across regions
  • Fails via Portal, CLI, ARM/Bicep, VS Code
  • Foundry project/resource creation succeeds; only model deployments are blocked

What we checked

  • Model availability ✔
  • Subscription & billing active ✔
  • IAM/RBAC correct ✔
  • No quota exhaustion ✔

Suspected cause

Based on Microsoft Q&A reports, error 715‑123420 indicates a platform‑level block by automated risk / fraud protection (RTFP) during deployment validation.

Request

Please review the subscription for RTFP / security blocking and unblock model deployments if applicable, or advise on required remediation steps.

Foundry Control Plane
Foundry Control Plane

Eine einheitliche Verwaltungsschnittstelle, die Sichtbarkeit, Governance und Kontrolle für KI-Agenten, -Modelle und -Tools im gesamten Foundry-Unternehmen bietet


1 Antwort

Sortieren nach: Am hilfreichsten
  1. Karnam Venkata Rajeswari 3,665 Zuverlässigkeitspunkte Externe Microsoft-Mitarbeiter Moderator
    2026-04-25T16:42:35.6733333+00:00

    Hello Nico Schröder

    Welcome to Microsoft Q&A .Thank you for reaching out to us.

    Thank you for the detailed case information and for completing extensive validation steps. The behavior described - where all model deployments consistently fail with error 715-123420 across regions, tools, and models - clearly indicates that the issue is not related to configuration, quota, permissions, or resource setup.

    Based on the observed pattern, the failure is occurring during backend service-level validation, before the model deployment is provisioned.

    This type of behavior is commonly associated with platform-level validation controls, which may include automated security or risk-based checks applied at the subscription level. When such controls are triggered, deployment operations can be uniformly restricted.

    We request you to please raise a support request for further assistance from the backend team to unblock the deployments. It can be raised in the following way. Please note that there could be minor UI changes, but the overall process would remain similar-

    1. Please go to the Azure portal > Help + support > New support request.
    2. Then for the Issue type select “Technical” and then under Service select “Azure AI Foundry” (or “AI Services”).
    3. Under Problem type, please choose “Other service issue”
    4. In the description, include: • Error code: 715-123420 , please provide the correlation IDs with a note that "suspect an RTFP/fraud‐block and would like it lifted"
    5. Then please submit the case. The RTFP team will validate your subscription status and remove any automated risk hold.

    The support process enables internal teams to review backend validation systems and determine whether any restriction is in place, followed by appropriate remediation if applicable.

    At this stage, progress depends on backend review by the responsible service teams. Assistance will continue through the support channel to ensure the request is investigated and addressed appropriately.

    Meanwhile we request to please double-check that

    • No policy assignments at the management-group or subscription level are set to Deny on Foundry deployments (see “Control AI model deployment with built-in policies”).
    • No pending compliance or billing holds on your subscription.
    • A paid (not free/trial) subscription is being used—RTFP blocks sometimes target trial accounts.

    The following references might be helpful , please check them out

     

    Thank you

    Please 'Upvote'(Thumbs-up) and 'Accept' as answer if the response was helpful. This will be benefitting other community members who face the same issue.

    War diese Antwort hilfreich?


Ihre Antwort

Antworten können von Fragestellenden als „Angenommen“ und von Moderierenden als „Empfohlen“ gekennzeichnet werden, wodurch Benutzende wissen, dass diese Antwort das Problem des Fragestellenden gelöst hat.