AIF-C01 Practice Test

The AWS Certified AI Practitioner (AIF-C01) is a foundational-level AWS certification. The exam has 65 questions, runs 90 minutes, costs $100, and requires a passing score of 700/1000. CertSharp provides 500 exam-style AIF-C01 practice questions — 30 free with no credit card — each with a detailed explanation.

1. Question 2Security, Compliance, and Governance for AI SolutionsEasy

   How can you implement the principle of least privilege for an AI development team using AWS?

   • A

     Attach the AWS managed policy AmazonSageMakerFullAccess to the entire team for convenience

   • B

     Create IAM roles scoped to the specific SageMaker actions and S3 buckets each team member requires

   • C

     Use a single shared IAM user with broad permissions to simplify team credential management

   • D

     Grant S3 bucket-level access to the entire team and restrict SageMaker access by source IP address

   Show answer & explanation

   Correct answer: B — Create IAM roles scoped to the specific SageMaker actions and S3 buckets each team member requires

   Least privilege requires granting each team member only the permissions needed for their specific role. Creating IAM roles scoped to required SageMaker actions and S3 buckets limits the blast radius of any credential compromise. AmazonSageMakerFullAccess grants excessive permissions beyond what most team members need. A shared IAM user violates individual accountability principles. IP-based restrictions do not replace resource-level permission scoping.

2. Question 3Fundamentals of AI and MLEasy

   What is the benefit of using SageMaker Studio over individual SageMaker services?

   • A

     It provides a unified, web-based IDE that integrates all SageMaker tools in a single environment

   • B

     It automatically selects the optimal training algorithm and hyperparameters for each dataset

   • C

     It reduces infrastructure costs by running ML workloads on shared serverless compute

   • D

     It continuously monitors deployed models and automatically retrains them when data drift is detected

   Show answer & explanation

   Correct answer: A — It provides a unified, web-based IDE that integrates all SageMaker tools in a single environment

   SageMaker Studio is a web-based integrated development environment that consolidates all SageMaker capabilities - notebooks, experiments, training, pipelines, and deployment - in one interface. SageMaker Autopilot (not Studio) automatically selects algorithms and tunes hyperparameters. SageMaker does not run all workloads on shared serverless compute. SageMaker Model Monitor can detect data drift but does not automatically retrain models.

3. Question 4Fundamentals of GenAIEasy

   A financial services company wants to use a GenAI model that complies with SOC 2 and ensures data is processed entirely within a specific AWS Region. Which service feature should they prioritize?

   • A

     Provisioned throughput, which reserves dedicated inference capacity to ensure consistent response times

   • B

     Model evaluation, which benchmarks model accuracy on domain-specific compliance test datasets

   • C

     Fine-tuning with proprietary datasets to align model outputs with internal compliance policies

   • D

     Regional availability with compliance certifications and data residency guarantees

   Show answer & explanation

   Correct answer: D — Regional availability with compliance certifications and data residency guarantees

   Ensuring SOC 2 compliance and data residency requires selecting a model deployed in a region that holds the required compliance certifications and guarantees data does not leave the specified region. Provisioned throughput manages inference capacity, not regulatory certification. Model evaluation assesses accuracy against benchmarks but does not provide compliance certification. Fine-tuning adjusts model behavior but does not guarantee regional data residency or certification status.

4. Question 5Guidelines for Responsible AIEasy

   What is a potential legal risk of using generative AI without human oversight?

   • A

     Higher cloud inference costs when longer prompts are processed during peak demand periods

   • B

     Intellectual property infringement if the model reproduces copyrighted text or images in its output

   • C

     Increased response latency that degrades the customer experience during high-traffic windows

   • D

     Gradual model drift that lowers answer accuracy as underlying product data changes over time

   Show answer & explanation

   Correct answer: B — Intellectual property infringement if the model reproduces copyrighted text or images in its output

   Reproducing copyrighted material from training data without human oversight exposes the organization to intellectual property infringement claims, which is a legal risk. The other options describe operational and technical concerns - inference cost, response latency, and model drift - rather than legal exposure, so intellectual property infringement is the only legal risk listed.

5. Question 6Applications of Foundation ModelsEasy

   A regulated financial firm must demonstrate to auditors exactly which datasets were used to train a model and how they were transformed. Which practice directly supports this requirement?

   • A

     Compressing training datasets into smaller representations to improve model convergence and reduce storage

   • B

     Identifying and removing discriminatory patterns from training data before fine-tuning begins

   • C

     Parallelizing data preprocessing steps across multiple compute instances to reduce pipeline latency

   • D

     Maintaining data lineage that tracks data origin, transformations, and usage across the AI lifecycle

   Show answer & explanation

   Correct answer: D — Maintaining data lineage that tracks data origin, transformations, and usage across the AI lifecycle

   Data lineage records where data originated, how it was transformed at each stage, and how it was used across the AI lifecycle - enabling reproducibility, regulatory auditing, and debugging. Data compression optimizes storage and speed but provides no audit trail of data provenance. Bias removal improves fairness but does not document data origins or transformation history. Parallelizing preprocessing accelerates pipelines but does not track lineage.

6. Question 7Security, Compliance, and Governance for AI SolutionsMedium

   A company uses Amazon Bedrock to generate marketing content. They must ensure the content doesn't violate brand policy. Which feature should they implement?

   • A

     Increase the sampling temperature to generate more creative and varied marketing outputs

   • B

     Enable prompt caching to reuse the system prompt prefix and reduce per-request latency

   • C

     Configure provisioned throughput to reserve dedicated inference capacity for high-volume campaigns

   • D

     Configure Bedrock Guardrails with topic deny lists and content filters to block off-brand output

   Show answer & explanation

   Correct answer: D — Configure Bedrock Guardrails with topic deny lists and content filters to block off-brand output

   Amazon Bedrock Guardrails enforce content policies by blocking or filtering outputs that match denied topics or harmful content categories, ensuring brand compliance at the output level. Increasing temperature makes outputs more random, raising the risk of policy violations. Prompt caching improves latency but does not filter or constrain output content. Provisioned throughput manages inference capacity, not content compliance.

7. Question 8Fundamentals of AI and MLMedium

   When you use Amazon SageMaker to deploy a model behind an HTTPS endpoint, what is automatically managed by the service?

   • A

     The endpoint's TLS certificates and auto-scaling

   • B

     The quality and validity of the incoming inference input data

   • C

     The business-level success metrics for the application

   • D

     The model's prediction explainability and feature attributions

   Show answer & explanation

   Correct answer: A — The endpoint's TLS certificates and auto-scaling

   SageMaker hosting automatically manages TLS certificate provisioning, endpoint auto-scaling, and the underlying compute infrastructure, so teams can focus on model development. Input data quality monitoring requires a separate SageMaker Data Quality monitor configuration. Business metrics are defined and tracked outside the endpoint infrastructure. Model explainability requires SageMaker Clarify configuration - it is not automatic.

8. Question 9Fundamentals of GenAIMedium

   A software development team wants AI-powered inline code suggestions, security vulnerability scanning, and an AI chat assistant integrated into their existing IDEs. Which AWS service provides these capabilities?

   • A

     Amazon Q Developer, which delivers inline code generation, security scans, and an in-IDE AI assistant

   • B

     Amazon SageMaker Studio, which provides a web-based ML development environment focused on training and experiments

   • C

     Amazon CodeGuru Reviewer, which performs asynchronous pull request analysis for code quality and security issues

   • D

     Amazon Kendra, which provides intelligent enterprise document search with no code generation or IDE integration

   Show answer & explanation

   Correct answer: A — Amazon Q Developer, which delivers inline code generation, security scans, and an in-IDE AI assistant

   Amazon Q Developer provides AI-powered inline code suggestions, real-time security vulnerability scanning, and an AI chat assistant directly within IDEs such as VS Code and JetBrains. SageMaker Studio is an ML development environment for training and deploying models, not for general developer code assistance. CodeGuru Reviewer analyzes code asynchronously on pull requests - it does not provide real-time inline suggestions. Amazon Kendra is an enterprise search service with no code generation capability.

9. Question 10Guidelines for Responsible AIMedium

   A company preparing for an AI audit must demonstrate its model's intended use, training data sources, and performance across demographic groups. Which SageMaker feature should they use?

   • A

     SageMaker Model Cards, which document intended use, training data, and per-group performance metrics

   • B

     SageMaker Model Monitor, which tracks data distribution changes and alerts when baselines are exceeded

   • C

     SageMaker Autopilot, which automates model selection and hyperparameter tuning across candidates

   • D

     Model quantization, which compresses model weights to reduce inference latency and deployment cost

   Show answer & explanation

   Correct answer: A — SageMaker Model Cards, which document intended use, training data, and per-group performance metrics

   SageMaker Model Cards are governance artifacts that document a model's purpose, intended use, training data characteristics, evaluation results, and performance across subgroups - providing the evidence required for responsible AI audits. Option B describes SageMaker Model Monitor. Option C describes SageMaker Autopilot. Option D describes model optimization techniques (quantization) unrelated to governance documentation.