Guest Gist: Everything Everywhere All at Once: Crime and Justice in the Age of Total Surveillance

In the classic detective story, a rumpled investigator paces a crime scene, notebook in hand, squinting at a cigarette butt or a smudged fingerprint. Columbo tugs at his raincoat and asks one more question. Miss Marple notices something out of place in a flower arrangement. The detective's gift is observation, intuition, the patient gathering of clues that coalesce into a narrative of guilt.

That world is vanishing. In its place rises something stranger and more powerful: a system where the clues already exist, captured automatically by a million cameras, sensors, and transaction logs, waiting only for the right query to surface them. The detective of the future may never leave her desk. The crime may be solved before the perpetrator even knows they're a suspect—or perhaps even before the crime itself actually occurs.

This is not science fiction. The infrastructure is falling into place. The question we face is not whether such a system is possible, but what kind of society it will create. Will the abolition of undetected crime be liberating, freeing us from the constant low-level anxiety of theft, fraud, and violence? Or will it prove suffocating, a world where every misstep is recorded and every minor transgression punished?

The Surveilled Society

Every face is now a registration plate.

Modern facial recognition systems have reached a level of accuracy that makes meaningful anonymity in public spaces nearly impossible. A face captured on a single camera can be matched against databases containing millions of images, linking a person to a name, a government identifier, an address, a history. The technology is good enough, and human faces are unique enough, that tracking an individual's movements through an urban area has become a matter of computational power rather than investigative effort.

But facial recognition is only one thread in a much denser web. It does not even need to be highly accurate, because it can be cross-referenced against other data streams that are far more reliable. Every journey on public transport leaves a record—an Oyster card tap, a Metro swipe, a ticket purchase. Every car has a registration plate, and those plates are photographed thousands of times per day by traffic cameras, parking systems, and automatic number plate recognition networks. Credit card transactions create timestamped, geo-located records of purchases. Loyalty cards track buying habits. Electronic toll systems log highway journeys to the meter.

Then there is the mobile phone, the remarkable device that most of us carry at all times. It is simultaneously a communication tool, a social media portal, a web browser, a payment wallet, and a GPS tracker. It knows where you are, who you're talking to, what you're reading, what you're buying, and increasingly, through fitness trackers and health apps, your heart rate and sleep patterns. The phone doesn't just generate data—it serves as the skeleton key that links all the other data streams together, connecting your face to your credit card to your transport card to your car to your social media accounts.

Online interactions add another dimension. Private messages on WhatsApp or iMessage may be encrypted end-to-end, but metadata—who contacted whom, when, for how long—remains visible to service providers and, under the right legal frameworks, to authorities. Public posts on social media are archived and searchable. Email, despite feeling ephemeral, persists on servers indefinitely.

The result is that it is now technically possible to construct a detailed timeline of nearly every movement, action, transaction, and communication in a person's life. Not just where they went, but who they met, what they bought, what they said, what they read, what they liked, what made them pause and scroll back up the page.

The surveilled society is not a future dystopia. It is the present, waiting only for someone with the technology and the legal authority to run the query.

Detection by Database Query

The classic detective worked forward from evidence. A crime occurred; the detective walked the ground , laid out the timeline and gathered clues; those clues led to suspects; investigation narrowed the field until the guilty party was identified. The process was laborious, uncertain, and dependent on human skill and intuition. Most crimes went unsolved because the evidence was simply not available, or because connecting the dots required more time than investigators could spare.

The new model inverts this. The evidence already exists, captured and stored automatically. Detection becomes less about gathering clues and more about querying databases—finding the signal in the noise of billions of recorded data points.

Consider a simple example: geofencing. When a crime occurs at a particular location and time, investigators can request data from mobile phone carriers showing every device that was present within a defined radius during a defined window. Instantly, they have a list of potential witnesses and suspects. Cross-reference that list against the victim's own phone records, social media connections, and transaction history, and patterns begin to emerge. Who was nearby? Who knew the victim? Who might have recently been in conflict with them?

These techniques are already in routine use. Investigators solving serious crimes regularly request tower dumps (records of all phones connected to a particular cell tower), analyse social network graphs to map relationships, and track suspects' movements through accumulated location data. Financial crimes are detected by analysing patterns in transaction data that would be invisible to any human auditor.

The shift is profound. Detection is no longer primarily about finding evidence that might not exist—it's about filtering through evidence that definitely does exist to construct a coherent narrative. The database of the everything and the everywhere is the haystack. The haystack contains the needles of specific evidence about a crime. The challenge is knowing which needles to look for, and how the needles are connected.

The Detective Who Never Sleeps: Artificial Intelligence and Crime

The database-query model of detection has an obvious limitation: scale. There are too many databases, too many records, too many crimes, and far too few investigators. A homicide might justify dozens of detectives spending months analysing data. A bicycle theft will not. The practical result is that most crimes, even those where evidence almost certainly exists somewhere in the data, remain unsolved simply because no one has time to look.

This is where artificial intelligence enters the picture.

The tasks that consume investigative time—analysing data, finding patterns, correlating events across multiple sources, constructing explanatory narratives—are the tasks that AI systems have become remarkably good at. A modern large language model can read thousands of documents and summarise them coherently. Computer vision systems can scan hundreds of hours of video footage and flag moments of interest. Graph analysis algorithms can map complex networks of relationships and identify suspicious clusters.

More significantly, AI systems can operate continuously, processing new data as it arrives, running queries that would take human analysts weeks in a matter of seconds. They do not get tired. They do not get distracted. They do not have a caseload that forces them to prioritise. 

The logical endpoint is an investigative system that operates with minimal human intervention. Crimes are reported—or detected automatically through anomaly detection in the data—and AI systems immediately begin querying the available databases, identifying suspects, and constructing narratives. It can do this all at once, for one or many crimes, reported or unreported. Human investigators review the AI's conclusions, verify the chain of reasoning, and make decisions about prosecution. The detective who never sleeps.

This is not a distant fantasy. Predictive policing systems already attempt to forecast where crimes are likely to occur. Fraud detection algorithms flag suspicious transactions in real time. The extension to full investigative AI is a matter of degree, not kind.

When Evidence and Detection Are No Longer Barriers, Which Crimes Should We Enforce?

Legal systems typically distinguish between "serious" crime and everything else. In practice, this distinction often determines which crimes receive investigative resources and which are effectively ignored. Murder gets a task force. Shoplifting gets a shrug.

But calling some crimes "unserious" is a misnomer. There is no such thing as a victimless minor crime in any meaningful sense. People whose cars are vandalised, whose packages are stolen, whose gardens are trampled by trespassers, whose phones are snatched—these people suffer real harm. They lose money, time, peace of mind. They feel violated and vulnerable. They lose trust in their communities and institutions.

The cumulative effect of unchecked minor crime is corrosive. A neighbourhood overwhelmed by fly-tipping, graffiti, antisocial behaviour, and petty theft becomes a neighbourhood where people lose confidence. If the authorities can't or won't address these problems, why should residents support them? Why should they take the law seriously at all? The social contract frays.

Consider what changes when investigative capacity is no longer the limiting factor. Minor crimes happen in plain sight all the time. The perpetrators walk past CCTV cameras. They use public transport to travel to and from the scene. They pay for materials—spray paint, alcohol, tools—with traceable transactions. They carry mobile phones that log their location. Many are foolish enough to film themselves and post the evidence on social media.

Today, this evidence goes largely unexamined. Tomorrow, an AI system could flag the incident, identify the perpetrators, and generate a case file—all automatically. The practical barriers that currently shield petty criminals would dissolve.

The question then becomes: should we pursue all these newly solvable crimes?

The case for doing so is pretty obvious. Laws exist for some good reason. If behaviour is harmful enough to criminalise, it should be harmful enough to enforce. Selective enforcement is inherently arbitrary and often discriminatory. A system that actually enforces all laws equally would be, in a sense, fairer than our current system of discretionary prosecution.

But the implications are troubling. Many laws are over-broad, outdated, or unjust. They remain on the books but selective enforcement leaves them dormant. Total enforcement would reveal the full scope of state power—and the full reach of behaviours that are technically criminal. It would also create a society where basically everyone is guilty of something. What would that mean for decisions on employment, immigration or housing?

Detection Is Not the Same as Enforcement

Suppose we can detect every crime. What then?

Our systems for detecting crime are imperfect but improving rapidly. Our systems for responding to detected crime—arrest, prosecution, adjudication, punishment—are imperfect and show no signs of improving at all. Courts have large backlogs, and prisons are overcrowded. Probation services are, generally speaking, overwhelmed. What happens when the caseload multiplies by a factor of ten? Or a hundred?

The options are unpalatable. Mass incarceration is expensive, socially destructive, and often counterproductive—people who go to prison tend to emerge more likely to reoffend, not less. Fines fall disproportionately on the poor. Community service programmes require administration and supervision that may not be available. Electronic monitoring extends the surveillance state into the home.

More fundamentally, what is the purpose of criminal punishment? Is it supposed to deter others? Rehabilitate the offenders? Lock them up where they can’t do more damage? Retribution for their sins? Each theory implies different responses to different crimes, and none offers clear guidance for a world where detection is near-universal.

There is also the question of legitimacy. Punishment is tolerated when it feels proportionate and fairly administered. A system that punishes every infraction, no matter how minor, may generate more resentment than compliance. At some point, enforcement becomes harassment, and law becomes tyranny.

Will Crime Be Futile? Is That a Good Thing?

The dream of the perfect crime has animated fiction and reality alike. From the locked-room mystery to the untraceable cryptocurrency heist, there is something that fascinates about the possibility of wrongdoing that leaves no trail. 

Total surveillance threatens to make perfect crime impossible. Every action leaves a trace, every trace is recorded and every record is searchable. The criminal who successfully evades all detection would need to operate entirely outside the digital world—no phone, no card, no car, no face recognisable by any camera. Such a person would be conspicuous by their very absence from the data.

On one level, this is straightforwardly good. Most crime causes real harm to real people, and a society with less crime is better than a society with more. If the certainty of detection prevents crimes from being committed in the first place, victims are spared suffering and perpetrators are spared the consequences of their choices.

Of course the analysis is not so simple. Not all laws are just, and not all enforcement is legitimate. Throughout history, people have broken laws to resist oppression, protect the vulnerable, expose wrongdoing, and challenge unjust systems. Fugitive slaves were criminals. So were those who hid Jews from the Nazis. Whistleblowers, protesters, dissidents—all have relied on some degree of undetectability to do what they believed was right.

A world without the possibility of undetected dissent is a world where state power faces no friction. Even in democracies, this is dangerous. The capacity to act outside official channels, to organise without surveillance, to maintain private spaces—these are not merely conveniences. They are preconditions for the meaningful exercise of freedom.

The futility of crime may also have subtler costs. A society that eliminates all risk of transgression may lose vitality. The possibility of rule-breaking, even when we choose not to break rules, creates a space for autonomy and self-definition. We are moral agents in part because we could do wrong but choose not to. In a world where the choice is removed, are we still free?

National Security, Intelligence, and Policing: Different Things, or Basically the Same Thing?

Historically, the institutions that protect societies from external threats have been distinct from those that maintain internal order. Intelligence agencies spy on foreign powers. Military forces defend borders. Police enforce domestic law. The separation is not merely administrative—it reflects different legal frameworks, different oversight mechanisms, and different relationships between the pursuer and the suspect.

Total surveillance erodes these distinctions. The same cameras that track suspected terrorists also track traffic violations. The same data analytics that identify foreign influence operations also flag tax fraud. The same AI systems that monitor social media for extremist content also detect ordinary criminality.

When the tools converge, there is a strong likelihood that the institutions will also converge. Police departments acquire advanced intelligence capabilities. Intelligence agencies expand into domestic surveillance. Data flows between agencies that were once separate, creating comprehensive profiles that no single institution could have built alone.

This convergence raises deep questions about power and accountability. Intelligence agencies have traditionally operated with far less oversight than police, justified by the secrecy necessary for national security. If those same agencies are effectively policing domestic populations, should they be subject to the same constraints as ordinary law enforcement? If police are using intelligence-grade surveillance, should they require warrants, be subject to judicial review, and transparent to the public they are supposed to serve?

The risk is a system that combines the secrecy of intelligence with the coercive power of policing, answerable to neither the courts nor the public. Such a system might be effective—terrifyingly so—but it could not be called the rule of law as we understood it in the 20th century and previously..

The Bargain We Have Not Yet Made

We are drifting toward a world of total surveillance without ever having consciously chosen it. No legislature debated and approved the comprehensive monitoring of public life. No referendum asked citizens whether they accepted the abolition of anonymity. The system emerged piecemeal, each component introduced for its own reasonable-seeming purpose—preventing terrorism, reducing fraud, managing traffic, improving customer service—until the whole became something no one intended.

It is still possible to make choices. We could decide that some data should not be collected, or should be deleted after defined periods, or should be inaccessible to authorities without judicial approval. We could distinguish between data collected for operational purposes and data retained for investigative purposes. We could insist that AI systems used in law enforcement be transparent, auditable, and subject to meaningful human review.

We could also decide, consciously and democratically, to embrace comprehensive surveillance as the price of comprehensive security. Such a choice would be neither irrational nor unprecedented—many societies have traded liberty for safety in times of perceived crisis. The important thing is that the choice be made openly, with full understanding of its implications, rather than by default.

What we should not do is avoid the question. The technology exists and will continue to improve. The data is being collected whether we acknowledge it or not. The AI systems are being developed, by governments and corporations alike. The only question is whether we will be their masters or their subjects.

Conclusion: The World We Are Building

The future of crime and justice is being written now, in code and silicon and fibre-optic cable, by engineers and executives and officials who may not fully grasp what they are creating. It is a future where privacy is vestigial, where anonymity is suspicious, where the possibility of undetected wrongdoing approaches zero.

This world may well have genuine benefits. People may be safer from many kinds of crime. Fraud may be harder to commit and easier to remedy. Children may be protected from predators who once operated in the shadows. The corrupt and the cruel may find fewer places to hide.

But it will also be a world fundamentally different from any that humans have previously inhabited. Throughout history, the practical impossibility of total surveillance created breathing room—space for eccentricity, dissent, experimentation, privacy. That space was not a design flaw to be corrected. It was the condition for human freedom, creativity, and dignity.

Now the system sees everything, everywhere, all at once. The question is no longer whether it can. The question is what we will ask it to do—and whether, in answering that question, we will remain the kind of society worth protecting.

This is not a technical question, and technologists cannot answer it for us. It is a political question, a philosophical question, ultimately a question about what kind of people we want to be. The machines are waiting for our instructions. It remains to be seen whether we will program them wisely.