SageMaker models should have network isolation enabled

How to fix it

Set at creation: fixing it means replacing the resource.

Turn network isolation on for the model, so no network call runs in either direction between its inference container and anything else and the container gets no AWS credentials. Check what the container does at run time first: anything that fetches code, a model artifact, a package or a licence check while serving stops working, while the artifact download and output write SageMaker performs itself, outside the container, keep working. There is no update operation for a model, so this means a new model with every endpoint, batch transform job and pipeline step repointed at it. SageMaker requires a model and its endpoint to agree on network isolation, so plan the endpoint move with it.

In the console

SageMaker Studio → Models → Deployable models → Create deployable model → Security → Network isolation

With the AWS CLI

aws sagemaker create-model --model-name <model-name> --execution-role-arn <execution-role-arn> --primary-container '{"Image":"<inference-image-uri>"}' --enable-network-isolation --region <region>

With Terraform

resource "aws_sagemaker_model" "this" {
  name                     = var.model_name
  execution_role_arn       = var.execution_role_arn
  enable_network_isolation = true

  primary_container {
    image = var.inference_image_uri
  }
}