Home / AIF-C01 / Objective 4.1

Domain 4: Guidelines for Responsible AI · 14% of the exam · Full exam outline

4.1 Explain the development of AI systems that are responsible

What this objective covers · 7% of the exam

Identify features of responsible AI (for example, bias, fairness, inclusivity, robustness, safety, veracity).; Explain how to use tools to identify features of responsible AI (for example, Amazon Bedrock Guardrails).; Define responsible practices to select a model (for example, environmental considerations, sustainability).; Identify legal risks of working with generative AI (GenAI) (for example, intellectual property infringement claims, biased model outputs, loss of customer trust, end user risk, hallucinations).; Identify characteristics of datasets (for example, inclusivity, diversity, curated data sources, balanced datasets).; Describe effects of bias and variance (for example, effects on demographic groups, inaccuracy, overfitting, underfitting).; Describe tools to detect and monitor bias, trustworthiness, and truthfulness (for example, analyzing label quality, human audits, subgroup analysis).

Practice questions

Question 1easy

A health insurer's member assistant on Amazon Bedrock must not engage in conversations about specific medical treatment advice. The team can describe that subject in plain language. Which Amazon Bedrock Guardrails safeguard fits this requirement?

Included in Sample: sample every domain.See plans

Choose an answer to check it. Afterward, select any option to review its explanation.

Word filters block exact custom words or phrases; a subject described in plain language can be phrased many ways, which a fixed word list cannot cover.

Grounding checks catch responses not supported by a source or irrelevant to the query; a well-grounded answer about treatment would still pass.

Denied topics let you define subjects that are undesirable in your application, and the guardrail blocks them when they appear in user queries or model responses.

Content filters detect predefined harmful categories such as hate or violence; treatment advice is not one of those categories.

Question 2easy

While reviewing a large language model it hosts, a team finds that the model's weights and activations are stored at a higher numeric precision than the task needs. Which optimization addresses exactly this to shrink the model and speed up inference with minimal effect on performance?

Included in Master: master every domain.See plans