Azure and the Bing API Shutdown - No Notice, No API
Azure removed Bing API resources, forcing users to switch to a restricted search tool within the AI Agents Service under stricter terms and data policies.
Table of Contents

Update 2025-03-11
Literally it keep's getting worse from day to day. Yeah Azure now enabled the creation for old subscriptions that already deployed a Bing-Search before. New Subscriptions are still left behind.
Update 2025-03-10
The creation of the Bing API resource is back. The same way it disappeared, it returned. Good that the support told you to switch to the proprietary Azure Agent SDK that is in beta and the whole Azure Agents which is in preview. Unannounced takedowns and literally no communication for more than a month ... ðŸ«
An official statement is yet to be found. Anybody who used the Bing Search for LLM had officially two days notice (while the API creation was long time before disabled ...) to switch to the Azure Agent SDK.

Introduction
The Bing API removal on Azure should be a bigger story. Minimal notice and no alternatives — exactly what you would expect from an enterprise cloud service.
The Bing API Shutdown
From one day to the next, Azure decided to remove the Bing Azure API. Without ANY notice period, they disabled the creation of a new API resource in Azure. Good luck if you want to move your resources now. The Microsoft Learn Q&A, Reddit etc. are full of people asking why they are getting an error in the form of:
[{"code":"ApiSetDisabledForCreation","message":"Your account is not approved for this resource. Contact BingAPIMS@microsoft.com with any questions."}]
Here are some links to the discussions in Microsoft Learn Q&A: 1, 2, 3, 4
And the worse part is that there is no official statement from Microsoft. The only thing you can find is a Retirement notice for the BingSearchAPIs with Your LLM on March 6, 2025. Yeah right, the retirement already happened in February as no new resources can be created! The notice was "Added to roadmap: 03/04/2025". This means there was officially only a 2-day notice prior to shutdown, and counting from when they quietly deactivated the ability to create new resources, approximately a month or more of unexplained service degradation before any official announcement.
To make it clear a notice for the standard API (the one without "your LLM") was never published! Or at least I could not find it.
The "Alternative" Offered
The provided alternative? Just use Grounding with Bing Search. Nope, this is not an alternative! Sure, there are workarounds, but they are workarounds... This resource can only be used as a tool in the Azure AI Agents Service. 💩
You now have to use their Azure Agent service to be able to use any websearch at all ... . Btw. the Agent Framework is still in preview and the npm package does not even have a stable version! So be prepared for random and breaking changes. Are enterprise customers really happy shipping alpha/beta and preview-releases?
Terms of Service Restrictions
According to their TOS:
you must only use Grounding with Bing Search with the AI Models permitted within the Azure AI Agent Service.
must only use Grounding with Bing Search with the AI Models permitted within the Azure AI Agent Service.
Don't even think of using non-permitted models! The TOS also include several statements about not using the grounded answers to train your own models.
Data and Privacy Considerations
While data collection was an issue before, they now state more clearly:
Microsoft will use data you send to Grounding with Bing to improve Microsoft products and services. When you send personal data to this service, you are responsible for obtaining sufficient consent from the data subjects. The Data Protection Terms in the Online Services Terms do not apply to Grounding with Bing.
A Heads Up
Just a heads up—you might need to switch to alternative search options such as:
- brave
- DuckDuckGo
- Serper
- searchapi
- firecrawl_dev
all thanks to Azure's increasingly restrictive changes. But the positive side is that you can now implement proper search with actual content of the webpages instead of just the metadata.