Module 2: Responsible Generative AI

Module 2: Responsible Generative AI#

Module 1 gave you a capable model and the skills to prompt it. Module 2 asks the harder question: how do you know it is any good, and how do you deploy it without causing harm? Powerful models are also unpredictable ones. They hallucinate, absorb bias, leak data, and can be manipulated. Using them responsibly is not an optional extra; it is a core engineering competency.

This module builds responsible AI from the ground up: first how to evaluate models, then the foundations and dimensions of responsible AI as a practice, and finally concrete techniques to improve security and safety.

No.

Chapter

What you will learn

1

Chapter 1: Evaluating LLMs

Why evaluation is hard, metric- and dataset-based approaches, benchmarks, and evaluation on Amazon Bedrock.

2

Chapter 2: Foundations of Responsible AI

What responsible AI is, its dimensions, the design-build-operate lifecycle, and how to assess an application’s risk.

3

Chapter 3: Dimensions of Responsible AI

The eight dimensions in depth: privacy and security, robustness, veracity, fairness and safety, transparency and explainability, governance, and controllability.

4

Chapter 4: Improving Security and Safety

Jailbreaking and prompt injection, guardrails, watermarking, and debiasing.

The arc is deliberate: you cannot manage what you cannot measure (evaluation first), you cannot act responsibly without a shared framework (foundations and dimensions), and only then can you apply targeted defenses (security and safety). The labs in Module 2 Labs: Responsible AI in practice make the defenses concrete on Amazon Bedrock.