By William Leiss and Richard Smith

Publication details forthcoming. Expected publication date Spring 2027*

This book begins with the story of the emergence of artificial intelligence (AI) for the general reader, detailing some of its benefits as well as its problematic side. It then brings the well-established risk management approach to bear on the catastrophic risks associated with Frontier AI, the newest AI models—risks which have been identified by many AI professionals as well as by leading AI firms. These risks include undermining cybersecurity, the creation of new biological and chemical weapons, and the real possibility that newer AI models may evolve beyond their human creators’ ability to understand or control the resulting systems. The Director of the CIA referred to the newest AI models as “digital nuclear weapons.”

Canada and other countries use the risk management approach for managing catastrophic risks, for example, those which arise from using nuclear power. When confronted by new catastrophic risks, the risk management approach first asks: How likely are they to happen? And if they happen, how bad will the outcomes be? This approach tells us (a) how severe the levels of risks are and (b) whether those levels are acceptable in terms of public security. We conclude that the catastrophic risks associated with AI are clearly unacceptable. Stringent regulations by governments must be created urgently, requiring the AI industry to implement a global indefinite pause on Frontier AI and on the goal of creating a superintelligent machine: Stop AI! We call upon the Government of Canada to enact such regulations and join with other countries to create an international treaty on AI safety.

Table of Contents

Preface

Part One: Artificial Intelligence

• Introduction

• Chapter 1: The Good

• Chapter 2: The Bad

• Chapter 3: The Ugly

• Chapter 4: The Superhuman Machine

Part 2: Risk Policy and Analysis

• Chapter 5: The Risk-Based Approach
• Chapter 6: The Control Problem
• Chapter 7: Catastrophic and Existential Risks
• Chapter 8: Risk Estimation
• Chapter 9: Quantitative Risk Assessment for Loss of Control (By Michael G. Tyshenko)

Part Three: Governance and Risk Management

• Chapter 10: Regulation
• Chapter 11: Risk Mitigation Options
• Chapter 12: Heart of Darkness
• Chapter 13: Blaming the Victims
• Chapter 14: What Should the Public Do?
• Chapter 15: What Should Governments Do?
• Chapter 16: Risk Management Summary

References

Acknowledgements

How we used AI

Early in the process of writing this book we used a variety of commercial generative AI tools (NotebookLM and Gemini, ChatGPT, and Claude) along with a few open source/open weight models (Gemma, deepseek, kimi, qwen, and gpt-oss), mainly for experimenting with and learning how models work - nothing like running a model on your laptop to give you an appreciation for the energy they consume (and cooling they require)!

In a practical sense, generative AI was involved in our online research, especially as Gemini became more and more integrated into the Google search results. The configuration for this web site (built in Markdown for Hugo and deployed using Workers/Pages on Cloudflare) was also largely done by ChatGPT and Claude, with help from CloudFlare’s in-house AI tool and Docker’s AI tool.

Generative AI also provided essential help early in the project when a mistake resulted in the entire bibliography having to be redone in Chicago (as required by MQUP) and converted from APA (as formatted by one of the authors - whose initials are RKS). Ahem.