Formerly known as Azure AI Services or Azure Cognitive Services is a unified collection of prebuilt AI capabilities within the Microsoft Foundry platform
Hi,
Since DNS now resolves to private IPs, your private endpoint and DNS configuration are likely correct.
In this setup, the most common cause of a generic “networking error” is the client access path.
When public network access is disabled, the resource is reachable only from:
- A VM inside the same VNet
A peered VNet
A machine connected via VPN/ExpressRoute
If Azure Content Understanding Studio (or your browser/client) is running outside the VNet, traffic will not reach the private endpoint — even if client ID / secret are valid.
I experienced a very similar scenario while working with a Launchpad subscription demo environment. Private endpoints were correctly configured and DNS resolved properly, but the service was accessible only from the provisioned VMs inside the same VNet. External access failed with a generic networking error despite correct credentials. The root cause was network isolation, not authentication.
You can validate quickly from a VM inside the VNet:
nslookup <resource-name>.services.ai.azure.com
Expected:
Resolves to private IP
TcpTestSucceeded : True
If this works internally but not externally, the issue is network reachability.
Also verify:
Private endpoint connection status = Approved
NSG allows outbound 443
Browser DNS over HTTPS is not bypassing VNet DNS
With private endpoints, network reachability is validated before authentication — so credential errors won’t appear if traffic cannot reach the resource.
Hope this helps.