Opening Case: Troy, New York
A mother walking her newborn looked up at the end of her block and saw a camera with a solar panel on top of it.
She lived in Troy, New York. The town has fifty-two thousand people. She had never seen the camera before.
What she found, after she went home and looked it up, was that the device was an AI-assisted license plate reader manufactured by a company called Flock Safety. Flock cameras read every plate that passes them. They identify the vehicle's make, model, and color. They log dents, roof racks, window stickers, and, on at least some models, a dog in the bed of a truck. They feed the result into a national database the company offers to police departments. Four thousand communities across the United States have them installed.
The mother's name is Dierdre Shea. After she finished reading, she walked the rest of her block and counted. Then she walked a few more blocks. Then she started telling her neighbors.
What Shea did not know, when she found the first camera, was that twenty-six Flock cameras had been installed across Troy. None of them had been put before the city council. None of them had been put before the residents. The contract, valued at $156,000 over two years, had been signed by the police department alone.
The city's procurement rules require council approval for any contract over $35,000.
The mayor of Troy is Carmella Mantello. She is a Republican. The city council is led by President Sue Steele. She is a Democrat. Once the cameras came to light, Steele asked how a contract of that size had been signed without council review. The police department's response was that the procurement had followed normal procedure.
Deputy Police Chief Steven Barker told reporters the cameras had been installed following a string of drive-by shootings. The cameras, he said, were used in almost every investigation the detective bureau pursued. He noted that the data was deleted after thirty days, and that the department had paused its participation in Flock's national database in March, after the first wave of public concern.
In May, the council voted to defund the cameras.
Mantello declared a state of emergency.
A state of emergency in Troy is normally reserved for severe weather, blizzards, and infrastructure failure. The mayor invoked one to keep the cameras running. 'I will not put our city in jeopardy and take these cameras away,' she said. She accused the council of trying to defund the police.
The council sued the mayor.
The cameras stayed up.
A few blocks from where Shea had found the first camera, residents gathered outside City Hall. Their signs read It's surveillance, not safety.
The company that makes the cameras has a position on this. Flock Safety's chief information security officer, Chris Castaldo, issued a statement after the controversy began. At Flock, he said, the company believed safety and privacy should go hand in hand. The technology was built around transparency, accountability, and local control. The platform included safeguards like audit trails to help ensure accountability at every step.
What audit trails have revealed elsewhere, when independent journalists obtained them, is a different picture.
In May, the technology site 404 Media published records showing police departments across the country had used Flock cameras to conduct searches on behalf of Immigration and Customs Enforcement. The federal agency had no contract with Flock. Local officers ran the searches anyway.
Other records showed Texas sheriff's deputies running Flock searches across the entire country, including Spokane County in Washington State, to locate a woman who had sought an abortion. The county's response, after the search came to light, was to withdraw its cameras from the national database. Other departments were asked to do the same. Some did. Some did not.
In a third case, a police chief in Kansas was forced to resign after his department's audit logs showed him using the cameras to track the movements of his ex-girlfriend.
None of these uses are what the Troy Police Department signed up for. None of them are what Flock Safety publicly endorses. None of them required a vote.
The cameras in Troy have not produced any of those outcomes. They might, given time, produce any of them. Or none of them.
The relevant fact is not what they have produced.
It is who decided they would be there.
Nobody on the city council. Nobody at a public hearing. Nobody who voted.
Twenty-six cameras across a town of fifty-two thousand people, signed by a single department, kept running by a state of emergency invoked for a purpose state emergencies were not invented for. A mayor and a council in court against each other. Residents in the street with handmade signs. A company in Atlanta with statements about audit trails that have shown, elsewhere, what audit trails on this technology actually record.
A mother walking her newborn noticed something at the end of her block.
The cameras are still running.
Introduction: The Pattern You Are About to See Everywhere
What you have just read is the pattern. The cameras in Troy were installed by people who did not specifically choose to install them at that scale. The mayor invoking a state of emergency to keep them running was not specifically choosing surveillance over the council's authority. The council suing the mayor was not specifically choosing a constitutional confrontation. The residents in the street with handmade signs were not specifically choosing protest as their first option. Each actor was responding to what had already been installed by the time their decision reached them. The cameras, the contract, the procurement bypass, the legal framework, the public outcry. Nothing about the situation was anyone's strategy. It was what happened when an unusually powerful technology arrived faster than the systems that would normally govern its arrival, and the cost of the gap was paid by people who were not in the rooms where the decisions were made.
That pattern is the subject of this book. The argument it amounts to can be stated in a single sentence: when powerful technologies arrive faster than the institutions built to govern them, defaults get installed by absence rather than by choice, and the cost is paid by people who were not in the rooms where the decisions were made. That sentence is the book's thesis. Every chapter that follows is evidence for it, illustration of it, or examination of what it would take to reverse. The thesis is not original in its parts. The insight that governance lags technology is as old as governance. What this book argues is that the lag has a structure, that the structure is consistent across every domain AI is entering, and that the structure is what produces the costs the domain-level conversations keep failing to name.
The case for naming it matters. Named patterns have different fates than unnamed ones. The wrong defaults install themselves in the space between what is happening and what has been named. Once you see the pattern, you will see it everywhere. Once you can name it, you will see that the people absorbing its costs have been waiting for the name.
Put most plainly: a default is just governance without a named author. Something is deciding. No one has put their name to the decision. The distance between those two facts is where the cost collects.
Once you see it, you will see it everywhere. In the schools arguing about cheating while the framing of an entire generation's relationship with cognition gets installed by default. In the workforce being displaced by technology no one is governing. In the institutions debating whether to allow AI at work while their employees are already using it. In the platforms that have no obligation to introduce a learning tool as a learning tool. In the regulators debating whether existing law covers what is happening. In the executives reading mission statements that no longer describe the company they run.
The pattern is the same in every case. Something is being installed by absence. Nobody specifically chose what is being installed. The cost is being absorbed by the people who never got to choose.
This book is not a complaint about AI. It is not a complaint about institutions. It is not a complaint about the people running the platforms, the schools, the agencies, the companies. The pattern this book describes does not require any of them to be acting in bad faith.
Naming the pattern is the work of this book. Pretending to fill the gap is not.
What I bring to this argument is breadth rather than depth in any single domain. The pattern becomes visible from operating across roles, industries, and pay levels, not from mastery within any one of them. More than twenty-five years of doing exactly that. The work has spanned grocery store jobs and pharmacy counters, frontline technical support and large-scale operations management, business development alongside city, county, and state officials, multiple specializations across financial services, two years as a para-educator with severely impacted high school students, and the advisory firm I now run.
I have been the one making the small decisions a system quietly absorbed, and the one making the large decisions it held to account. I have lived inside the gap between what an institution said it valued and what it actually rewarded. I have seen a few organizations attempt a different approach on purpose. One real difference was consistent. It is the subject the book builds toward.
Three distinctions travel across every chapter that follows. The first is the difference between what is available and what is accessible. AI dissolves the barriers that locked people out of expert knowledge for centuries, but only for the people who are shown what the tool is actually for; for everyone else the barriers stay exactly where they were. The second is the difference between a decision and a default. A decision has an author who can be named and held to it. A default installs itself when no decision is made, and most of what this book describes was installed the second way. The third is the difference between who is in the room and who pays, because the two are almost never the same people, and the gap between them is where the cost of the whole pattern finally lands.
These three lenses are not separate arguments. They are different views of the same pattern, and the book uses them across every domain it examines.
This is not a book about whether AI is good or bad. That debate is happening, and this book is not in it. This is a book about a structural pattern that operates regardless of how the debate resolves. If the optimists are right and the technology turns out to be net positive, the pattern determines who absorbs the cost of getting there. If the pessimists are right and it turns out net negative, the pattern determines what the damage looks like in practice. Either way, the pattern decides.
The book is written for the reader who is watching this happen, who senses something is going wrong and has not yet had the language to name it. It is also for the reader who has the language but has not yet seen how far the same pattern reaches, across schools and workplaces and agencies and the ordinary middle of a working life. And it is for anyone who, having seen it, wants to know what is actually theirs to do about it, at whatever scale they occupy, whether or not the institutions ever reach the work.
The book ends with what it would take to install something different, and with the one piece of that which does not wait for anyone's permission. Between this introduction and that ending is the work of seeing the pattern in the places it most needs to be seen.