Mcrosoft Foundry GPT model deployment fails with unusual activity error

Open an Azure support request against the affected subscription and ask for a backend review of the Foundry/Azure OpenAI “unusual activity” deployment block; changing models, capacity, regions, or recreating resources does not clear this type of restriction.

Microsoft Q&A cases with the same message—including failures across multiple models and resources—required Microsoft’s backend team to review and unblock the subscription or resource.^1^ Similar cases were resolved only after Microsoft completed analysis and adjustments.^2^

Open the support request

1. In the Azure portal, open the affected Foundry/Azure OpenAI resource.
2. In its resource menu, select Support + Troubleshooting.
3. Complete the diagnostic prompts. Since automated troubleshooting has already failed, select Create a support request.
4. Choose the same Pay-As-You-Go subscription that owns the resource. The assigned engineer can investigate only the subscription selected in the request.
5. Select Technical, then choose the closest available service and problem type for Azure OpenAI, Microsoft Foundry, or model deployment.
6. Select the active Developer support plan and an appropriate severity.
7. In the description, explicitly request:

   Backend review and removal of any account-, subscription-, or resource-level risk restriction blocking Microsoft Foundry model deployments with the “unusual activity” error.

The official workflow documents the resource-level Support + Troubleshooting route, subscription selection, diagnostic-data option, support plan, severity, and contact method.^3^

Creating the request requires Owner, Contributor, Support Request Contributor, or a custom subscription-level role containing Microsoft.Support/*.^3^

Include the evidence needed for review

Provide these details without posting subscription or account identifiers publicly:

• Exact error message and any numeric error code, such as 715-123420, if displayed.
• UTC date and time of the latest failed attempt.
• Affected subscription ID, resource name, resource group, region, and Foundry project.
• Model names and versions attempted, including GPT-4.1-mini and GPT-5.1.
• Deployment SKU/type and requested capacity.
• Whether deployment was attempted through the Foundry portal, Azure portal, CLI, or another client.
• Request, correlation, trace, or operation ID from the failed deployment.
• Confirmation that different models and capacities failed and automated diagnostics did not resolve it.
• A screenshot or exported Activity Log event, with secrets and personal information removed.

Azure Activity Log records subscription-level resource operations and is collected automatically.^4^ In the resource, open Monitoring → Activity log, locate the failed deployment operation around the UTC timestamp, and copy its correlation ID and failure details.

While the case is reviewed

• Do not repeatedly retry deployments or delete and recreate resources. A matching case states that retries, region changes, and resource recreation did not clear the review block and cautions that repeated recreation can worsen the activity pattern.^5^
• No additional identity verification form should be completed unless the support engineer provides one. Account-specific verification or authorization instructions must come through the private support case.
• This message does not by itself prove abusive use. Foundry employs automated and, in some circumstances, human review to detect and assess potentially abusive usage patterns.^6^

After submission, monitor Help + support → All support requests. Open the case and select New message to add correlation IDs or requested evidence.^7^

References

1. Azure OpenAI: "unusual activity" BadRequest blocks all model deployments (scale + new) across my subscription - Microsoft Q&A
2. Error 715-123420 when deploying model in Foundry - Microsoft Q&A
3. Create an Azure support request
4. Monitor Azure OpenAI
5. Model Deployment error: 715-123420 - Microsoft Q&A
6. Abuse monitoring
7. Manage an Azure support request