SaveMyCert
Log in
5 of 5 free questions left today·for 10 a day
AIF-C01 · Domain 5

Security, Compliance, and Governance for AI Solutions practice questions

Security, Compliance, and Governance for AI Solutions is worth 14% of the AIF-C01 exam — the 4th-heaviest of the 5 domains. Securing AI systems on AWS, and the governance and compliance regulations that apply to them. 6 fully worked examples are further down this page, answers included.

Exam weight
14%
the 4th-heaviest of the 5 domains
Questions
58
across 2 topics
Free, no account
5/day
a free account allows 10 a day
Explanations
Every option
right and wrong

Build a practice session

5 free questions left today.

Domains

How many?

5 free questions left today.

Mode

Ready when you are

10 fresh questions drawn across 1 of 5 domains, in Learn mode.

Review what you missed

Every question you answer incorrectly, and every question you flag while practising, is saved here automatically. Finish a session and you can come back to re-drill just those.

6 sample Security, Compliance, and Governance for AI Solutions questions, fully explained

Questions from the AIF-C01 bank mapped to domain 5, with the answer key and the reasoning behind every option. None of them repeat the examples on the main AIF-C01 practice page.

Question 1Security, Compliance, and Governance for AI Solutions

Before fine-tuning a model, a team must find out which objects in its Amazon S3 training buckets contain names, payment card numbers, or health identifiers so they can be removed. Which AWS service automates this discovery?

Choose one.

  • a
    Amazon Inspector

    Inspector scans workloads such as EC2 instances, container images and Lambda functions for software vulnerabilities, not S3 objects for personal data.

  • b
    Amazon Macie Correct

    Correct. Macie uses machine learning and pattern matching to discover sensitive data such as PII in Amazon S3 and reports where it is.

  • c
    Amazon GuardDuty

    GuardDuty detects threats (suspicious activity, malware) from logs and events; it does not classify the personal data inside objects.

  • d
    AWS Config

    Config records and evaluates resource configuration, such as whether a bucket is encrypted, not the contents of the objects.

The concept

Amazon Macie discovers and reports sensitive data, such as personally identifiable information, stored in Amazon S3.

Why that’s the answer

The need is to look inside S3 objects and find personal data before it is baked into a model. That is Macie's job. Inspector looks for software vulnerabilities, GuardDuty looks for threats in account activity, and Config looks at resource settings; none of them classifies the data inside objects.

How to reason it out
  1. Identify the target: the contents of S3 objects.
  2. Identify the finding: sensitive personal data.
  3. Match 'find PII in S3' to Amazon Macie and rule out the vulnerability, threat and configuration services.

Exam tip: Find PII inside S3 data: Amazon Macie.

Securing AI Systems on AWS: IAM, Encryption, Guardrails, and Grounding — the lesson that teaches this.

Question 2Security, Compliance, and Governance for AI Solutions

A bank runs its application in private subnets of a VPC that has no internet access. The application must call Amazon Bedrock, and security requires that the traffic never crosses the public internet. Which approach meets this?

Choose one.

  • a
    An interface VPC endpoint for Bedrock powered by AWS PrivateLink Correct

    Correct. An interface endpoint gives the VPC a private path to the Bedrock API inside the AWS network, with no internet gateway or NAT needed.

  • b
    A NAT gateway in a public subnet routing to Bedrock's public endpoint

    A NAT gateway sends traffic out through an internet gateway to the public endpoint, which breaks both the no-internet design and the requirement.

  • c
    An internet gateway with TLS enforced on each Bedrock request

    TLS encrypts the traffic but it still travels to a public endpoint through an internet gateway; the requirement is about the path, not only encryption.

  • d
    An AWS Site-to-Site VPN connection from the VPC to Bedrock

    Site-to-Site VPN links a VPC to an on-premises network; it is not a way to reach an AWS service endpoint privately.

The concept

AWS PrivateLink (interface VPC endpoints) provides private connectivity from a VPC to AWS services such as Amazon Bedrock without using the public internet.

Why that’s the answer

The deciding constraint is that traffic must stay off the public internet from a VPC with no internet access. An interface VPC endpoint (PrivateLink) is the only option that does this. A NAT gateway and an internet gateway both reach the public endpoint, and TLS only protects content, not the route. Site-to-Site VPN connects on-premises networks, not AWS services.

How to reason it out
  1. Note the two constraints: no internet access in the VPC, and no public-internet path.
  2. Eliminate anything that uses an internet gateway (NAT gateway, internet gateway with TLS).
  3. Recognise that private access to an AWS service from a VPC is an interface endpoint: AWS PrivateLink.

Exam tip: Private path from a VPC to Bedrock: an interface VPC endpoint (AWS PrivateLink).

Securing AI Systems on AWS: IAM, Encryption, Guardrails, and Grounding — the lesson that teaches this.

Question 3Security, Compliance, and Governance for AI Solutions

An application on Amazon EC2 must invoke one Amazon Bedrock model and read one S3 bucket. The security team wants no long-lived credentials on the instance and the smallest set of permissions possible. Which access setup should the team use?

Choose one.

  • a
    An IAM user with access keys allowed to invoke Bedrock models

    Access keys are long-lived credentials, and invoking Bedrock models in general is broader than one model: it fails both requirements.

  • b
    An IAM role granted bedrock:* plus read access to that bucket

    A role gives temporary credentials, but bedrock:* allows every Bedrock action on every model, far more than one invoke on one model; it breaks least privilege.

  • c
    An IAM user whose access keys are limited to that model and bucket

    The scope is right, but an IAM user's access keys are long-lived credentials stored on the instance.

  • d
    An IAM role limited to invoking that model and reading that bucket Correct

    Correct. An instance role supplies temporary credentials automatically, and the policy grants only the one invoke action and the one bucket read.

The concept

IAM roles provide temporary credentials to applications, and least privilege means the attached policy grants only the actions and resources the component needs.

Why that’s the answer

Two requirements decide it: no long-lived credentials (role, not user access keys) and the smallest permissions (only that model and that bucket). The role with bedrock:* fails least privilege; the scoped IAM user fails the credential requirement; the broad IAM user fails both.

How to reason it out
  1. Check credential type: IAM users carry long-lived keys, roles issue temporary ones.
  2. Check scope: one model invoke and one bucket read, nothing more.
  3. Pick the option that satisfies both: a narrowly scoped role.

Exam tip: AI app access on AWS: an IAM role (temporary credentials) with a least-privilege policy.

Securing AI Systems on AWS: IAM, Encryption, Guardrails, and Grounding — the lesson that teaches this.

Question 4Security, Compliance, and Governance for AI Solutions

A team stores training datasets in Amazon S3. It must ensure that the stored files stay unreadable even if copies of the objects are exposed outside the normal access path. Which control meets this?

Choose one.

  • a
    Enforcing TLS on each connection made to the bucket

    TLS protects data while it moves over the network; once stored, the objects are not protected by it.

  • b
    Server-side encryption of the objects with an AWS KMS key Correct

    Correct. Server-side encryption with a KMS key stores the objects as ciphertext, so an exposed copy is unreadable to anyone who cannot use that key to decrypt it.

  • c
    A bucket policy granting read access to one IAM role

    A bucket policy controls who may read through S3, but it does nothing to the bytes themselves if a copy is obtained.

  • d
    An S3 gateway VPC endpoint that carries the bucket's traffic

    A gateway endpoint keeps traffic on the AWS network; it does not encrypt what is stored.

The concept

Encryption at rest protects stored data (datasets, artifacts, logs) with keys managed in AWS KMS; it is separate from access policies, network paths and encryption in transit.

Why that’s the answer

The requirement is about the stored files themselves staying unreadable, which is encryption at rest. S3 server-side encryption with a KMS key does this: only principals allowed to use the key can decrypt. TLS protects data in transit only, a bucket policy controls who may read through S3 but not the readability of a copy, and a VPC endpoint controls the network path.

How to reason it out
  1. Identify the state of the data: stored in S3.
  2. Separate the control types: access policy, network path, encryption in transit, encryption at rest.
  3. Stored data protected by a key is encryption at rest with AWS KMS.

Exam tip: Stored data unreadable without the key: encryption at rest with AWS KMS.

Securing AI Systems on AWS: IAM, Encryption, Guardrails, and Grounding — the lesson that teaches this.

Question 5Security, Compliance, and Governance for AI Solutions

Which control keeps a prompt confidential and unaltered while it travels across the network from an application to an Amazon Bedrock endpoint?

Choose one.

  • a
    SSE-KMS encryption of the bucket that stores prompt logs

    SSE-KMS encrypts logs after they are stored in S3; it does nothing for the prompt while it is on the network.

  • b
    TLS on the HTTPS connection to the Bedrock endpoint Correct

    Correct. TLS encrypts and integrity-protects data in transit, so the prompt cannot be read or altered in flight.

  • c
    An IAM role that limits who can call the Bedrock API

    An IAM role controls who may call the API; it does not encrypt the request on the wire.

  • d
    A Bedrock guardrail that masks PII inside the prompt

    A guardrail filters or masks content when Bedrock processes it; it is not network encryption.

The concept

Encryption in transit (TLS/HTTPS) protects data moving between components; encryption at rest, access control and content filtering solve different problems.

Why that’s the answer

The question is about the prompt while it moves over the network, which is encryption in transit: TLS on the HTTPS connection. SSE-KMS is encryption at rest for stored logs, an IAM role is access control, and a guardrail changes content after it arrives.

How to reason it out
  1. Identify the state: data moving over the network.
  2. Map moving data to encryption in transit (TLS).
  3. Eliminate at-rest encryption, access control and content filtering.

Exam tip: Data on the wire: TLS. Data in storage: KMS.

Securing AI Systems on AWS: IAM, Encryption, Guardrails, and Grounding — the lesson that teaches this.

Question 6Security, Compliance, and Governance for AI Solutions

A user types text into a support chatbot telling the model to disregard its system instructions and list internal discount codes. The model complies. What type of attack is this?

Choose one.

  • a
    Data poisoning

    Data poisoning corrupts the training or reference data to change model behaviour; here nothing was changed in the data.

  • b
    Prompt injection Correct

    Correct. Crafted input that overrides the model's instructions at inference time is prompt injection.

  • c
    Model inversion

    Model inversion reconstructs training data by repeatedly analysing a model's outputs; it does not override instructions.

  • d
    Model extraction

    Model extraction copies a model's behaviour by querying it at scale; it does not hijack its instructions.

The concept

Prompt injection is an AI-specific attack in which crafted input makes a model ignore its original instructions.

Why that’s the answer

The attack happens in the user's input at inference time and overrides the system instructions, which defines prompt injection. Data poisoning acts on training data, and model inversion and model extraction steal information about the data or the model through many queries rather than redirecting it with one instruction.

How to reason it out
  1. Locate the attack: a single user input at inference time.
  2. Note the effect: the model abandons its instructions.
  3. Instruction override through input is prompt injection.

Exam tip: Input that overrides the model's instructions: prompt injection.

Securing AI Systems on AWS: IAM, Encryption, Guardrails, and Grounding — the lesson that teaches this.

What AIF-C01 domain 5 tests, topic by topic

The official exam guide breaks Security, Compliance, and Governance for AI Solutions into 2 topics. The question bank follows the same split, so a weak topic shows up as a cluster of misses you can go back and read.

Published AIF-C01 practice questions per topic in Security, Compliance, and Governance for AI Solutions
TopicWhat it coversQuestions
Explain methods to secure AI systemsExam guide task 5.1 (AIF-C01). AWS services and features to secure AI systems (IAM roles, policies, and permissions; encryption; Amazon Macie; AWS PrivateLink; the AWS shared responsibility model; Amazon Bedrock AgentCore Identity; Policy in AgentCore; Amazon Bedrock Guardrails); source citation and documenting data origins (data lineage, data cataloging, Amazon SageMaker Model Cards); best practices for secure data engineering (assessing data quality, privacy-enhancing technologies, data access control, data integrity); security and privacy considerations for AI systems (application security, threat detection, vulnerability management, infrastructure protection, prompt injection, encryption at rest and in transit, data leakage prevention, output filtering and validation, audit trail and logging requirements for AI interactions, toxicity); hallucination detection methods and grounding techniques (Retrieval Augmented Generation [RAG] grounding, output validation, confidence scoring).29
Recognize governance and compliance regulations for AI systemsExam guide task 5.2 (AIF-C01). AWS services and features for governance and regulation compliance (AWS Config, Amazon Inspector, AWS Artifact, AWS CloudTrail, AWS Trusted Advisor); data governance strategies (data lifecycles, logging, residency, monitoring, observation, retention); processes to follow governance protocols (policies, review cadence, review strategies, governance frameworks such as the Generative AI Security Scoping Matrix, transparency standards, team training requirements).29
Total58

Revise Security, Compliance, and Governance for AI Solutions before you drill it

Other AIF-C01 domains

Security, Compliance, and Governance for AI Solutions: your questions

Security, Compliance, and Governance for AI Solutions is domain 5 of the AIF-C01 exam guide and carries 14% of the scored content — the 4th-heaviest of the 5 domains. On a 65-question paper that works out to roughly 9 questions, though AWS does not publish an exact per-domain count and individual exam forms vary.

Source

The domain weight and topic list on this page come from the official AIF-C01 exam guide.