← All learning paths

Learning path · 5 steps

Ship an Azure AI workload safely

Design explicit identity and network boundaries, test untrusted input, and make a release decision with evidence.

Cloud engineers & application teams

  1. 01

    Practical guide

    Start Azure AI access with workload identity

    Select the correct runtime identity, API and role scope.

    Open guide →
  2. 02

    Practical guide

    Verify Azure AI network isolation end to end

    Verify inbound, outbound and dependency network paths.

    Open guide →
  3. 03

    Practical guide

    Build a prompt-injection defense test plan

    Exercise untrusted-input and unauthorized-action scenarios.

    Open guide →
  4. 04

    Practical guide

    Build an AI telemetry baseline you can investigate

    Confirm that expected logs reach the investigation workspace.

    Open guide →
  5. 05

    Practical guide

    Use an evidence-based AI release checklist

    Record evidence, exceptions and a rollback owner before release.

    Open guide →