Bitzu

AIF-C01 ยท Securing and governing AI solutions

Securing an AI workload on AWS

IAM roles and policies, encryption with AWS KMS, Amazon Macie, AWS PrivateLink, the shared responsibility model, Bedrock AgentCore Identity and Policy, and Bedrock Guardrails.

Key points

  • IAM roles and least-privilege policies: who and what can call models and data.
  • AWS KMS: keys that encrypt data.
  • Amazon Macie: finds sensitive data in S3.
  • AWS PrivateLink: keeps traffic off the public internet.
  • Shared responsibility: AWS secures the cloud itself; you secure identities, data and configuration.
  • Bedrock Guardrails: filter prompts and responses.
  • AgentCore Identity and Policy: govern what agents may do.

Source: https://docs.aws.amazon.com/bedrock/latest/userguide/guardrails.html

Practice questions

Single choice

Which service uses machine learning and pattern matching to discover sensitive data such as PII in Amazon S3?

  1. Amazon Macie
  2. AWS Config
  3. AWS Artifact
  4. AWS CloudTrail
Show answer

Answer: A. Amazon Macie

Amazon Macie scans S3 objects for sensitive data and raises findings. Config tracks resource configuration, Artifact hands out compliance reports and CloudTrail records API activity.

Select all that apply

A company runs a chatbot on Amazon Bedrock. Which tasks stay the customer's responsibility under the shared responsibility model?

  1. Granting least-privilege IAM permissions to callers
  2. Configuring guardrails for the chatbot
  3. Choosing and managing encryption keys for its data
  4. Physical security of the AWS data centers
  5. Patching the hosts that run the Bedrock service
Show answer

Answer: A. Granting least-privilege IAM permissions to callers, B. Configuring guardrails for the chatbot, C. Choosing and managing encryption keys for its data

Identities, configuration and data protection are yours. Physical facilities and the managed service hosts are AWS's side of the model.

True or false

True or false: Bedrock Guardrails can only evaluate text that is sent to a foundation model through Bedrock.

Show answer

Answer: False

False. The ApplyGuardrail API evaluates text on its own, without invoking a model, so it can protect content produced elsewhere.

Practice 8 questions on this topic in Bitzu

Short lessons, instant explanations and spaced review of what you get wrong. Free on iOS and Android.

Download on the App StoreGet it on Google Play

More AIF-C01 topics in this module