Generative AI Leader Practice Question: An e-commerce company is using a generative AI…

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

• ✗

  Cloud DLP (Data Loss Prevention)

  Why it's wrong here

  Cloud DLP discovers and redacts sensitive data such as card numbers and personal identifiers in storage and streams; it inspects content for privacy, not for hate speech or violence. It is tempting because it also filters model input and output, but its classifiers target data sensitivity, so it cannot satisfy this safety requirement.

• ✗

  Vertex AI Explainable AI

  Why it's wrong here

  Explainable AI attributes which input features drove a prediction, supporting debugging and compliance reporting. It reports why a model produced output, not whether that output contains hate speech or violence, so it cannot block harmful generations. It is tempting because it also inspects model behaviour, but its purpose is interpretability, not content safety filtering.

• ✓

  Vertex AI Safety Filters

  Why this is correct

  Vertex AI Safety Filters apply configurable thresholds that block harmful categories—hate speech, violence, harassment—directly on model input and output, satisfying the requirement to prevent harmful product descriptions. Configuring these filters enforces content policy at inference time, which is precisely the constraint the stem demands.

• ✗

  Vertex AI Model Monitoring

  Why it's wrong here

  Vertex AI Model Monitoring tracks serving data for drift, skew and anomalies in deployed models; it does not inspect or filter generated text for hate speech or violence. It would be correct for detecting input distribution shifts, whereas content safety requires configuring safety filters or thresholds on the generative model itself.

Go deeper

Related to this question