AWS Certified Generative AI Developer - Professional (AIP-C01) Domain 3

AI Safety, Security, and Governance

Official Exam Guide: Domain 3: AI Safety, Security, and Governance

Skill Builder: AWS Certified Generative AI Developer - Professional Exam Prep

---

Domain Overview

Domain 3 (20% of exam) focuses on implementing input/output safety controls, data security and privacy, AI governance and compliance, and responsible AI principles.

---

Task 3.1: Implement input and output safety controls

Key Skills:

• Develop comprehensive content safety systems (filter harmful inputs)
• Create content safety frameworks (prevent harmful outputs)
• Develop accuracy verification systems (reduce hallucinations)
• Create defense-in-depth safety systems
• Implement advanced threat detection (prompt injection, jailbreak)

Essential Documentation:

• Amazon Bedrock Guardrails
• Guardrail Components
• Amazon Comprehend
• Grounding Responses with Knowledge Bases

---

Task 3.2: Implement data security and privacy controls

Key Skills:

• Develop protected AI environments (VPC, IAM, Lake Formation)
• Develop privacy-preserving systems (PII detection, anonymization)
• Create privacy-focused AI systems (data masking, retention policies)

Essential Documentation:

• Amazon Bedrock VPC Endpoints
• Detect PII with Amazon Comprehend
• Amazon Macie
• AWS Lake Formation

---

Task 3.3: Implement AI governance and compliance mechanisms

Key Skills:

• Develop compliance frameworks (model cards, data lineage, audit logs)
• Implement data source tracking
• Create organizational governance systems
• Implement continuous monitoring and governance controls

Essential Documentation:

• SageMaker Model Cards
• AWS Glue Data Catalog
• AWS CloudTrail
• Amazon Bedrock Model Invocation Logging

---

Task 3.4: Implement responsible AI principles

Key Skills:

• Develop transparent AI systems (explanations, confidence metrics)
• Apply fairness evaluations
• Develop policy-compliant AI systems

Essential Documentation:

• Amazon Bedrock Agent Trace
• Amazon Bedrock Model Evaluation
• AWS Responsible AI

---

AWS Service FAQs

• Amazon Bedrock FAQs
• Amazon Macie FAQs
• Amazon Comprehend FAQs
• AWS CloudTrail FAQs

---

Study Tips

1. Master Amazon Bedrock Guardrails - Guardrails implement content filtering, PII redaction, topic denial, and word filters. Essential for safety.

2. Understand hallucination reduction - Learn grounding with Knowledge Bases, confidence scoring, fact-checking patterns, and structured outputs.

3. Learn PII detection and protection - Master Amazon Comprehend PII detection, Macie for discovery, and Bedrock native privacy features.

4. Study governance requirements - Understand model cards, data lineage tracking, audit logging, and compliance frameworks.

5. Practice threat detection - Learn to detect and prevent prompt injection, jailbreak attempts, and adversarial inputs.

---

Note: This is Domain 3 of 5, representing 20% of exam content.