Governance

What this chapter covers

Governance is 8% of the Databricks Generative AI Engineer Associate exam, the smallest chapter. It covers how you keep a generative AI application safe, legal and compliant while it still performs well.

The chapter has one topic: masking sensitive data as a guardrail, guardrail techniques against malicious user input such as prompt injection, the legal and licensing rules for data sources, and better alternatives when a data source contains problematic text.

Questions are mostly short scenarios in which the stated requirement (latency, cost, freshness, who may access what, quality signal) decides the answer. They are single-answer and some multiple-answer multiple choice, and short Python or SQL is shown as text. There are no exhibits.

How the topics fit together

Governance starts with the data: only sources you may legally use, cleaned and masked before they are indexed. At run time, guardrails on the serving endpoint check what users send and what the model returns.

• Masking questions ask how to hide sensitive values without blocking useful answers.
• Malicious input questions ask which defense stops prompt injection or jailbreak attempts.
• Licensing questions ask whether a source or model may be used for the stated purpose.
• Mitigation questions ask for the better fix when a source contains toxic, biased or private text.

How to use these pages

Each topic page has the same shape, so you can skim for what you need.

• What it is: the idea in plain words.
• How it works: the mechanism, with tables and examples.
• What to remember for the exam: the facts to memorize and how questions are phrased.
• Common mistakes: the classic traps.
• Check yourself: three questions to test your understanding.