GitHub · GH-600
Autonomy levels and risk
GH-600 · Lesson 15 of 16 · 2:04:05
YouTube loads only after you press play. This chapter opens at its timestamp in the full course. Watch on YouTube ↗
In this lesson
Prepare for the GH-600 exam (GitHub Certified: Agentic AI Developer). In this lesson: classifying agent actions by operational, security, and compliance risk, the four autonomy levels (read-only, propose-only, execute with guardrails, human-authorized execution), assigning levels for delivery speed and compliance with a machine-readable risk level and progressive autonomy, responsible AI standards, and the short list of decisions that need human judgment. Includes three original practice questions.
Preview the topics
- Contoso’s triage agent reads new issues and writes a summary with a suggested owner in its job summary, but it never changes the repository. Which autonomy level fits?
- Contoso wants agents to merge dependency patch updates without waiting for a person, while staying compliant. What makes this acceptable?
- Which of these Contoso agent actions should require human judgment?