AIF-C01 Guidelines for Responsible AI Practice Question

Which THREE considerations are important when implementing responsible AI for a production NLP system? (Choose three.)

⚠ Common exam trap

AWS often tests the distinction between general security practices (like encryption) and the specific pillars of responsible AI (fairness, transparency, accountability, and robustness), leading candidates to confuse data protection with ethical AI governance.

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

✓

Continuously monitor model outputs for bias and drift

Option B is correct because responsible AI in production requires ongoing monitoring of model outputs to detect bias and data/concept drift, since model behavior can degrade or become unfair as real-world data changes over time. Option D is correct because model cards are a standard transparency artifact that document a model's intended use, performance metrics across relevant subgroups, and known limitations, enabling informed and accountable use by stakeholders. Option E is correct because embedding bias detection as an automated gate in the CI/CD pipeline ensures every model update is evaluated for fairness regressions before deployment, making responsible AI a repeatable part of the ML lifecycle rather than an afterthought. Option A does not belong because FDA approval applies to regulated medical devices and clinical AI, not to general production NLP systems. Option C does not belong because encrypting training code at rest is a generic security control and is not a responsible-AI consideration such as fairness, transparency, or accountability.

Answer analysis

Option-by-option breakdown

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

✗
Obtain FDA approval for the model
Why it's wrong here
FDA approval governs medical devices and diagnostics, so it is irrelevant to a general production NLP system and is not a responsible-AI consideration. It is tempting because regulatory approval signals accountability, and it would be correct for a clinical decision-support model subject to medical device regulation.
✓
Continuously monitor model outputs for bias and drift
Why this is correct
Production NLP models degrade as language, data and user behaviour shift, and bias can emerge or amplify post-deployment. Continuous monitoring of outputs for bias and drift satisfies the responsible-AI requirement by detecting harmful changes early, enabling remediation before affected users suffer sustained harm.
✗
Apply encryption at rest for all training code
Why it's wrong here
Encrypting training code at rest is a general security control, not a responsible-AI consideration such as fairness, transparency, or bias mitigation for NLP outputs. It is tempting because encryption is a genuine best practise, and it would be correct in a question about data protection or infrastructure hardening.
✓
Publish model cards detailing intended use, performance, and limitations
Why this is correct
Model cards document intended use, performance across subgroups, and known limitations, giving downstream users the context to judge fitness for purpose. Publishing them satisfies the transparency and accountability requirement of responsible AI by preventing misuse and uninformed reliance on the NLP system.
✓
Include bias detection in the CI/CD pipeline for every model update
Why this is correct
Continuous bias detection embedded in the CI/CD pipeline catches fairness regressions at every model update, before deployment. This satisfies the responsible AI requirement for ongoing monitoring rather than one-off audits, ensuring production NLP outputs remain equitable as data and weights change.

Go deeper

AIF-C01 Guidelines for Responsible AI Practice Question

Why each option matters

Option-by-option breakdown

Related to this question