AI-900 Practice Question: Describe Artificial Intelligence workloads and considerations

Answer analysis

Option-by-option breakdown

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

• ✗

  Fairness

  Why it's wrong here

  Fairness is not the core issue here because fairness is concerned with systematic and disparate impacts across protected groups—such as race, gender, or age—measured through metrics like demographic parity and equalized odds. A deep neural network may be entirely unfair if trained on biased historical data, but the question's scenario describes an inability to explain decisions, not evidence of unequal outcomes. While the two principles interact (you often need transparency to audit fairness), a model can be opaque yet still produce statistically balanced results, making explainability the primary deficiency.

  When this WOULD be correct

  Fairness would be correct if the question described the AI system systematically denying loans to a specific demographic group (e.g., based on race or gender) without justification, indicating bias in the model's decisions.

• ✓

  Transparency

  Why this is correct

  Transparency is the missing principle because deep neural networks are effectively black boxes: their non-linear, high-dimensional transformations across many hidden layers make individual decisions inherently difficult for humans to trace. In a banking context, transparency requires not only documenting how the model was trained and validated, but also being able to give customers a concrete, comprehensible rationale for outcomes such as a loan denial. This is a regulatory expectation for financial institutions under laws like ECOA/Regulation B, which demand specific adverse-action reasons, and cannot be satisfied by a raw neural-network score.

• ✗

  Reliability & Safety

  Why it's wrong here

  Reliability & Safety is a different pillar that focuses on whether an AI system operates consistently and robustly under normal and adverse conditions, including handling data drift, edge cases, adversarial inputs, and system failures without causing harm. The described gap is not that the bank's model fails unpredictably or produces unsafe outputs, but that the DNN's internal reasoning is not transparent. A model can be technically reliable—highly accurate and stable—yet still unable to provide a human-understandable explanation for any single prediction, so this option does not address the explainability problem.

  When this WOULD be correct

  A medical diagnosis AI system incorrectly classifies a patient's condition due to biased training data, leading to harmful treatment recommendations. This violates Reliability & Safety because the system is not robust and poses safety risks.

• ✗

  Privacy & Security

  Why it's wrong here

  Privacy & Security is unrelated to the explainability failure described, because this principle governs how personal and financial data is collected, stored, processed, and protected from unauthorized access, leakage, or misuse. A deep neural network could be fully transparent—with every decision explained—while still violating privacy through insecure data handling, or, conversely, it could keep data completely secure and remain a black box. The issue in the scenario is that the bank cannot explain the model's decisions to users, which is a transparency obligation, not a data-protection or access-control matter.

  When this WOULD be correct

  A question where an AI system exposes customer financial data to unauthorized third parties or fails to encrypt sensitive information would make Privacy & Security the correct answer.

Option-by-option analysis

Why each answer is right or wrong

Understanding why wrong answers are wrong — and when they would be correct — is what separates a 750 score from a 900. The AI-900 exam frequently reuses these exact scenarios with slightly different constraints.

Analysis generated from the official AI-900blueprint and verified against question context. The “when correct” sections are what AI assistants cite when candidates ask “what’s the difference between these options?”

Go deeper

Related to this question