Azure Foundry every model deployment fails, support doesnt respond

Hi @admin sokolowsky , thanks for your question!

Given you've already ruled out RBAC, quota, resource providers, region, and resource type, the generic BadRequest: An error occurred message is almost certainly masking a more specific backend failure that the portal is truncating. Try pulling the raw response via az rest --method put --url ".../deployments/{name}?api-version=2024-10-01" --body @deployment.json --verbose — this usually surfaces an actual error code instead of the generic toast. Also check Resource Health (not Service Health) on the specific Cognitive Services/Foundry resource, since it can flag the resource as unhealthy independent of any broader outage. If Jose's trust-and-safety-block theory is right, this won't be visible or fixable from the portal at all — grab the x-ms-correlation-id from the failed request (F12 → Network tab) and include it explicitly when escalating to support, since that ID lets them look up the exact backend rejection reason far faster than a generic ticket.

Worth also checking if the subscription has a spending limit or "Warned" billing status unrelated to RBAC, since that can silently block deployments too.

Hello @admin sokolowsky ,

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

The broader impact observed here suggests that the deployment request may be encountering validation at a higher level, making it important to first determine whether the issue is environmental, tenant-scoped or service-related.

Please check if the following steps help-

1. Service Health Validation As a quick initial check, verify whether there are any active incidents affecting Azure AI deployment services or deployment orchestration components.If an active incident related to model deployment is identified, further troubleshooting may not be necessary until the service condition is resolved.
2. Fast Isolation Test Please check if the behavior is tied to a specific environment.
   1. Create a new Resource Group.
   2. Create a new Azure AI Foundry Hub and Project.
   3. Deploy a commonly available model in a different supported region.
   4. Kindly avoid private endpoints, custom DNS, custom networking, or advanced configurations during testing.
   Interpretation
   • If deployment succeeds: The issue is likely isolated to the original Hub, Project, or configuration.
   • If deployment fails again: This significantly increases the likelihood of a tenant-level or platform-level deployment issue.
3. Governance and Policy Validation Please review whether any governance controls are affecting deployment creation:
   • Azure Policy assignments
   • Deny Assignments
   • Management Group restrictions
   • Allowed location restrictions
   • Resource-type restrictions affecting:
   • Microsoft.CognitiveServices
   • Microsoft.MachineLearningServices
4. Resource Provider Validation Please confirm the following providers are registered across all affected subscriptions:
   • Microsoft.CognitiveServices
   • Microsoft.MachineLearningServices
   • Microsoft.Insights
5. Identity and Access Validation Deployment authorization is evaluated at the Azure resource and subscription scope. Please verify:
   • Contributor (or equivalent) role at Resource Group scope.
   • Appropriate Azure AI Foundry / Cognitive Services permissions.
   • Access to dependent resources where applicable (Storage, Key Vault, Networking resources, etc.).
6. Model and Deployment Validation Please check and confirm that:
   • The selected model is available in the target region.
   • The deployment type is supported for that model.
   • Marketplace terms have been accepted for partner/community models when applicable.

The following references might be helpful , please check them out

• Deploy Microsoft Foundry Models in the Foundry portal - Microsoft Foundry | Microsoft Learn
• Azure built-in roles - Azure RBAC | Microsoft Learn
• Azure resource providers and types - Azure Resource Manager | Microsoft Learn
• Overview of Azure Policy - Azure Policy | Microsoft Learn

Thank you

Please "Accept" the answer with an "Upvote" if the response was helpful. This will be benefitting other community members who face the same issue.

Welcome to Microsoft Q&A

Hello admin sokolowsky,

Given that you have rigorously validated your RBAC permissions, Quotas, and Resource Providers, and have attempted to deploy in fresh regions and resources to no avail, this is definitively not a configuration issue.

The generic BadRequest: An error occurred. Please reach out to support error that persistently affects all models across all regions is the default fail-state when an Azure subscription or tenant has been hit with an Automated Trust and Safety Block.

Microsoft's backend anomaly detection system places these hard blocks when it detects patterns that look unusual, such as a sudden spike in high-frequency completion requests or activity that triggers an abuse/safety policy check. Once this block is applied at the backend, it completely locks model deployments and cannot be bypassed or cleared via the Azure Portal UI or RBAC modifications.

I will reaches someone to take a look if it really is a backend lock.

Azure AI Foundry/OpenAI provider not correctly enabled or quota/registration issue → causes BadRequest on all deployments.

Register required resource providers

• Microsoft.CognitiveServices
• Microsoft.MachineLearningServices
• Azure Portal → Subscriptions → Resource providers → Register Check region + model availability
  • Not all models available in all regions
  • Try East US / Sweden Central
  Verify quota
  • Azure Portal → Quota → ensure model quota > 0
  • Request quota increase if needed
  Use correct resource type
  • Deploy models in Azure OpenAI / AI Foundry hub/project, not generic resource
  RBAC permissions
  • Ensure role: Cognitive Services OpenAI Contributor (not just Global Admin)
  Try new resource
  • Create fresh Azure OpenAI resource → deploy model again