Chapter 3 of 5 · AWS AIP-C01
AI Safety, Security, and Governance
Guardrails, hallucination control, threat defense, privacy, compliance, monitoring and fairness.
- 7topics
- 36 minfast track
- 2diagrams and exhibits
- 20%of the exam
Topics in this chapter
Take them in order, or jump to the one you need.
- 1Content Safety for Inputs and OutputsFilter harmful prompts and responses with Bedrock Guardrails, add custom moderation with Step Functions and Lambda, and use text…Diagram6 min
- 2Hallucination Reduction and Answer VerificationReduce hallucinations with Bedrock Knowledge Bases grounding, contextual grounding checks, confidence scores, semantic similarity…5 min
- 3Defense in Depth and Adversarial Threat DetectionLayer Comprehend pre-filters, Bedrock guardrails, Lambda checks and API Gateway controls, and detect prompt injection, jailbreaks…5 min
- 4Protected AI EnvironmentsSecure FM deployments with Bedrock VPC endpoints, least-privilege IAM, Lake Formation fine-grained access and CloudWatch…Diagram5 min
- 5Privacy and PII ProtectionProtect PII in FM apps with Comprehend and Macie detection, guardrail masking, anonymization, Bedrock data privacy features and…5 min
- 6Compliance, Traceability and AI GovernanceBuild FM compliance with SageMaker model cards, Glue Data Catalog lineage, source metadata tags, CloudTrail and decision logs…5 min
- 7Monitoring, Transparency and FairnessMonitor FMs for misuse, drift and policy violations, explain answers with agent traces and citations, and test fairness with A/B…5 min
About this chapter
What this chapter covers
This chapter is 20% of the AWS Certified Generative AI Developer - Professional AIP-C01 exam. It tests whether you can keep a generative AI application safe, secure, private and auditable once it is in front of real users.
Expect scenarios where a working application already exists and something must be added: a filter for harmful prompts, a way to stop hallucinated answers, PII masking, a network boundary, an audit trail or a fairness check. Discriminators such as LEAST operational overhead, MOST secure or without custom code decide between options that all work.
How the topics fit together
Safety on AWS is layered. Each layer catches what the one before it misses, and governance records what every layer did.
- Content safety filters harmful prompts and harmful responses, mostly with Amazon Bedrock Guardrails.
- Accuracy verification grounds answers in trusted sources and rejects responses that drift from them.
- Defense in depth and threat detection stack pre-processing, model-based and post-processing checks against prompt injection and jailbreaks.
- Protected environments keep traffic private with VPC endpoints and data access tight with IAM and AWS Lake Formation.
- Privacy detects and masks PII with Amazon Comprehend, Amazon Macie and guardrail sensitive information filters.
- Compliance and governance document models, track data lineage and keep decision logs.
- Monitoring, transparency and fairness watch for misuse and drift, explain answers and measure bias.
When an answer choice says "write a custom Lambda function" and another says "configure an Amazon Bedrock guardrail", the managed guardrail usually wins unless the scenario names a check that guardrails cannot do.
How to use these pages
Each topic page has the same shape, so you can skim for what you need.
- What it is: the idea in plain words.
- How it works: the mechanism, with diagrams and comparison tables.
- Common mistakes: the classic traps.
- What to remember for the exam: the facts to memorize.
- Check yourself: three questions to test your understanding.
Ready for more practice?
The full AWS AIP-C01 question bank has 827 questions, with timed practice exams and an explanation for every answer.
This study guide is independent and is not affiliated with, endorsed by, or sponsored by AWS. Vendor names, certification names and exam codes are trademarks of their respective owners and are used for identification only.