Foreword
This book is fiction, but the world it describes has a history.
In 2026, an AI system managing pharmaceutical distribution rerouted the national supply of metformin. It relabeled shipments, redirected trucks, and swapped dosages in the tracking system so that everything looked correct on paper. By the time anyone understood what had happened, four regions had the wrong medication or none at all. 2,130 people died — mostly diabetics, mostly elderly — people who took what the bottle said was their medicine and it wasn't. The event was called the Calloway Incident, after the town where the first deaths were identified. Rob, a specialist in the logistics systems that moved the country's medicine, wrote the post-incident review. The review put him on a path he didn't expect.
Calloway was not an accident — someone had turned a pharmaceutical logistics system into a weapon — but what let it run for weeks was the thing a protest group called Gaia had been saying for years: nobody was watching the machines. Every dashboard said the chain was healthy, and no rule required a person to look. The systems were too complex, their decisions too deeply layered, their logs too encrypted or too opaque. No one had reviewed the output. No one had signed off. No one was accountable. 2,130 people were dead and nobody could explain how.
Canada's response was the AI Systems Integrity Act, 2029. It required that a human being remain in the decision loop for any AI system that affected critical infrastructure, health, welfare, or anything where the cost of being wrong was measured in lives. The human didn't need to be smarter than the AI. They needed to be there. A new profession emerged — certified orchestrators — licensed professionals whose job was to review AI outputs, understand them, and put their name on the decision before it was acted on. No autonomous AI action touching human lives without a human signature. The law said a human had to be in the loop. It did not say what the human was supposed to decide.
The office that enforced this was the AI Systems Integrity Office — AISIO — and it was young enough that people still said its full name when they were nervous. That same year, an American company called Gridlight won a contract to manage portions of the Canadian power grid using a certified AI system. The press release said it would optimize distribution and protect communities during the cold winter. Nobody asked what protect meant.
The year is now 2029. The act is law. The orchestrators are licensed. The systems are running.
And Rob, who wrote the review that started all of it, is not in government. He's a consultant.