Generative AI Leader Google Cloud's Generative AI Offerings Practice Question
A bank's risk team must review every AI-generated customer communication for compliance before it is sent. They want to enforce a policy that blocks any message containing prohibited financial advice and routes flagged messages to a human reviewer. Which Vertex AI capability should they configure?
⚠ Common exam trap
The trap here is assuming that built-in safety filters cover organization-specific compliance policies like prohibited financial advice, when such policies require custom detection logic.
Answer choices
Why each option matters
Answer the question above first, then reveal the full breakdown to understand why each option is right or wrong.
Correct answer & explanation
✓
A custom classifier deployed on Vertex AI Endpoints and invoked in the generation workflow
The bank needs an inline, policy-specific gate on each generated message. A custom classifier trained on compliant and non-compliant examples, deployed to a Vertex AI Endpoint, can be called in the generation pipeline to score each message; the application then blocks or routes based on that score. Monitoring, standard safety filters, and offline evaluation do not provide real-time domain policy enforcement and routing.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✗
Vertex AI Model Monitoring with skew and drift detection
Why it's wrong here
Model Monitoring tracks statistical changes in input features and predictions over time, such as training-serving skew and prediction drift. It does not inspect the semantic content of generated text or enforce policy rules, so it cannot block messages containing prohibited financial advice or route them to reviewers.
- ✓
A custom classifier deployed on Vertex AI Endpoints and invoked in the generation workflow
Why this is correct
A custom classifier trained on examples of compliant and non-compliant communications can detect prohibited financial advice, and deploying it to a Vertex AI Endpoint lets the application call it after generation. The workflow can then block flagged text and route it to a human reviewer, satisfying the compliance policy.
- ✗
Vertex AI safety filters configured with adjustable harm thresholds
Why it's wrong here
Safety filters target categories such as harassment, hate speech, and dangerous content, and they can block responses that violate those categories. Prohibited financial advice is a domain-specific compliance rule, not a standard harm category, so safety filters alone would not reliably detect or route those messages.
- ✗
Vertex AI Model Evaluation with an automatic metric threshold
Why it's wrong here
Model Evaluation computes quality metrics such as BLEU or ROUGE against reference data during development or periodic assessment. It does not run inline on each generated message, cannot enforce a blocking policy in real time, and does not provide a routing mechanism to human reviewers.
Go deeper
Related to this question
About these practice questions
Courseiva writes every Generative AI Leader question from scratch — 1,008 in total, each with an explanation and a wrong-answer breakdown. None are copied from real exams or dumps. Learn why practice questions differ from exam dumps →
JA
Written and reviewed by Johnson Ajibi, MSc IT Security
Senior Network & Security Engineer · founder of Courseiva
Last reviewed September 2026 · checked against the official Google Cloud exam blueprint
This Generative AI Leader practice question is part of Courseiva's free Google Cloud certification practice question bank. Courseiva provides original exam-style practice questions with explanations, topic-based practice, mock exams, readiness tracking, and study analytics to help learners prepare for the Generative AI Leader exam.