An Azure service that provides access to OpenAI’s GPT-3 models with enterprise capabilities.
Hi @John Ellsworth welcome to the Microsoft Q&A community.
This specific error code (15-123420) is a hard security block on Azure OpenAI Services triggered by automated abuse prevention systems.
When you rapidly create, deploy, or tear down Azure OpenAI models or deployments during lab exercises, Microsoft’s automated risk engine flags the subscription for suspicious/bot-like quota exploitation.
What is Happening Behind the Scenes
- Service-Specific Risk Flag: Because this only impacts OpenAI models while your broader Azure subscription remains active, the flag is isolated to
Microsoft.CognitiveServices/ Azure OpenAI resource deployment APIs. - Support Ticket Blocking: When a subscription is placed under an automated security hold, the standard "Technical Support" ticket submission workflow is often restricted or redirected, which is why your ticket attempt failed.
- Paid Support Plan Won't Help immediately: Upgrading to Developer/Standard support will not automatically remove an abuse/security hold. Do not spend money upgrading your support plan for this issue, as security blocks are handled outside standard technical SLA queues.
Action Plan to Unblock Your Labs
Step 1: Submit a Subscription & Management Ticket (Free)
You do not need a paid support plan to reach Azure support for subscription, quota, or security block issues. Subscription and Billing management tickets are completely free for all Azure accounts.
- Go to the Azure Portal and search for Help + support.
- Click Create a support request.
- Set the following fields carefully:
- Issue type:
Subscription management(orService and subscription limits (quotas)) - Problem type:
Subscription AccessorOpenAI Resource Access / Security Block
- Issue type:
- In the description, clearly state:
"My account was flagged with Error 15-123420 while performing repeated Azure OpenAI model deployments for AI-102/AI-103 exam preparation labs. This is legitimate educational activity. I cannot open a technical ticket due to the hold, so please route this to the Cognitive Services / Fraud & Abuse Review team to clear the restriction on my subscription."
Step 2: Request a Review via Azure OpenAI Special Request
If the standard ticket portal gives you an error when submitting, submit the official Azure OpenAI Quota & Access Request:
- Navigate to the Azure Portal → Azure OpenAI Service.
- Click on Request Quota Increase or use the Azure OpenAI Limit Increase Request Form.
- Explaining that your deployments are automated lab scripts for certification prep helps human reviewers distinguish your usage from malicious API scraping.
Step 3: Workarounds to Continue Studying Immediately
While waiting for Microsoft support to remove the flag:
- Use Azure AI Studio / Resource Reuse: Instead of creating and deleting the Azure OpenAI resource and model deployments in every lab iteration, create one permanent Azure OpenAI resource and re-use the deployment endpoint across labs.
- Try a Different Region: If your lab script deploys resources to a single region (e.g.,
East US), manually change the target region in your template/script toSweden Central,West US 3, orFrance Central. Security flags are sometimes applied at the regional resource provider level. - Pass-Through via Standard OpenAI API: If the lab section strictly tests prompt engineering, REST calls, or SDK integration rather than Azure-specific ARM templates, you can temporarily swap in a standard platform OpenAI API key (
api.openai.com) to keep practicing without changing your code structure significantly.
N/B: I have generated the above answer using co-pilot as an AI tool. Also I have validated and updated the AI output.
I hope the information provided has been helpful to you! If so, please accept the answer by clicking the "Upvote and Accept Answer" on the post. We value your feedback, and it will help to assist others who might have a similar query.