Book
Seisei AI no Anzensei Nyumon (An Introduction to Generative AI Safety)
Koki Wataoka
Summary
A Japanese-language introduction to generative AI safety by Koki Wataoka, for engineers and researchers. Generative AI is transforming society and industry, but it has also surfaced risks such as harmful content and unintended behavior; the book systematically organizes those risks and how to address them. Chapter 1 defines AI safety and surveys recent trends. Chapter 2 classifies the threats faced in practice. Chapter 3 uses OpenAI's published model spec as a case study for thinking through ideal AI behavior. Chapter 4 introduces methods for evaluating safety, after which chapter 5 covers red-teaming. Chapters 6 and 7 explain techniques for improving safety at the model level and the system level, and the final chapter considers the future of AI safety. It brings together, in one book, the knowledge needed to develop and operate generative AI safely.
Target Readers
- Students and businesspeople who have become interested in AI safety
- Developers and researchers who work on AI safety professionally
Tags
Colophon
- Publisher
- Gijutsu-Hyohron
- ISBN
- 978-4-297-15702-9
- Published
- Jun 2026
- List price
- ¥3,300incl. taxMay differ from the actual selling price on Amazon
Get this book
* The link above is an advertisement via Amazon Associates.Related Books
Prerequisites
- Related
Web Security Tantosha no Tame no Zeijakusei Shindan Start Guide, 2nd Edition (A Vulnerability Assessment Start Guide for Web Security Staff)
Sen Ueno Teaches New Techniques for Preventing Information Leaks
Sen Ueno
Reason: After gaining hands-on diagnostic experience with tools like OWASP ZAP and Burp Suite that closes the gap between theory and practice, turn that same posture — verifying defenses in the field — toward a new target. Generative-AI risk is evaluated through a different form, benchmarks and red-teaming, but the idea of backing up desk assumptions with verification carries over.
Next Books
- Recommended
LLMOps
Managing Large Language Models in Production
Abi Aryan
Reason: Once you've learned generative-AI risk classification, red-teaming-based safety evaluation, and system-level countermeasures like guardrails, you advance to embedding them into a mechanism that actually keeps running in production. LLMOps' closing chapter covers responding to security risks such as governance, privacy, and prompt injection, plus building an observability pipeline — connecting the evaluation and countermeasure knowledge from the AI safety primer into concrete production-operation mechanisms.
- Related
Seisei AI Application Hyoka Nyumon (An Introduction to Evaluating Generative AI Applications)
Shinsuke Matsuki
Reason: After learning the techniques of safety evaluation via benchmarks and red-teaming, you broaden your view to how the act of 'evaluating' concretely gets built into your own generative-AI product. This book on generative-AI application evaluation covers everything from building an evaluation perspective model to confusion matrices, RAG-oriented metrics, security evaluation, and AI agent evaluation along the development lifecycle, connecting the abstract idea of safety evaluation to your own product's quality-assurance process.