The AI revolution is here, and most of the conversation is still focused on the same question:
Who loses their job?
That matters. Jobs are immediate, salaries are personal, and redundancy is easy to imagine.
But there is another group that may eventually have far more to fear from AI: the people who currently believe they control the system.
Politicians. Regulators. Large corporations. Wealthy individuals. Anyone whose position has historically depended on complexity, influence, selective enforcement, or simply being difficult to investigate.
Because AI may become very good at something we rarely discuss.
Looking backwards.
We Already Built the Surveillance System
Britain does not need to invent a surveillance society from scratch. Most of the infrastructure already exists.
Cameras watch roads and town centres. ANPR records vehicle movements. Banks maintain transaction histories. Phones generate location data. Retailers track purchases. Companies file accounts. Properties register owners. Tax records accumulate.
Almost every action in modern life produces a digital trail.
The limitation has largely been human. There are simply too many databases, transactions, companies and relationships for investigators to connect. A department may examine thousands of cases, but modern society generates billions of data points.
AI changes that.
The breakthrough is not necessarily collecting more information. It is understanding what is already there.
A simple way to imagine this is through what is already happening in business and government.
Businesses are rushing to incorporate AI, connect systems and optimise how they work. Many already have to keep detailed records for accounting, compliance and audit purposes. Local authorities are moving in the same direction, trying to reduce costs and improve efficiency.
Now imagine that a few years from now the government passes relatively simple legislation allowing more of those systems to communicate with one another.
Suddenly, the dots begin to connect.
From Ordinary Fraud to Powerful Networks
The first uses will probably be familiar: benefit fraud, undeclared income, false invoices, tax discrepancies and insurance claims.
AI can identify patterns such as spending that does not match declared income, repeated payments without explanation, or conflicting location data.
At first, I suspect many people may welcome this. The familiar argument will return:
If you have nothing to hide, you have nothing to fear.
But there is a huge difference now.
That principle has historically been directed downwards, towards the public.
Why would a mature investigative system stop there?
It would not care whether the subject was a plumber or a multinational owner, a council tenant or a parliamentarian.
It would not be intimidated by status, legal budgets or corporate structures.
It would simply look for patterns.
Money moving through connected companies. Ownership hidden across jurisdictions. Contracts linked to personal relationships. Wealth that cannot be explained by declared income.
Complexity has often protected powerful people because complexity consumes human time.
To an investigator, a network of offshore companies and intermediaries can be a maze.
To a machine, it is a graph.
That is the difference.
The machine may know exactly who you are.
It just may not care.
The Past Becomes Searchable
The most uncomfortable possibility for the people pushing these systems is that future AI will not only examine behaviour as it happens.
It will revisit the past.
Someone may commit wrongdoing today believing the trail is too obscure, too fragmented or too complicated ever to attract attention.
Years later, the records are still there: bank transactions, corporate filings, ownership histories, travel logs and property purchases.
What changes is the intelligence examining them.
Twenty apparently unrelated events in 2026 might look like one coherent pattern in 2033.
The evidence did not suddenly appear.
The ability to see it did.
We already have a version of this with DNA. Improvements in forensic technology continue to reopen old cases, correct miscarriages of justice and identify offenders years after the original crime.
AI could do something similar with financial and digital evidence.
A future investigation could reconstruct the money, people, companies, properties, dates and relationships behind an event that once seemed buried.
People may discover that they did not get away with something.
They merely acted before society had the tools to understand what it was seeing.
And once people understand that old actions can be re-examined by far more capable systems, time stops feeling like protection.
The Inversion of Status
Governments and large institutions will build these systems for understandable reasons: more tax collected, less fraud, lower costs and better enforcement.
Initially, ordinary people will probably face the greatest scrutiny.
But the obvious questions will follow.
If AI can identify undeclared income, why can it not identify suspicious corporate structures?
If it can find benefit fraud, why not procurement fraud?
If it can reconstruct a tradesman’s finances, why not a politician’s?
The people who build a system capable of looking at everyone may eventually find themselves included.
That could create a new form of equality before the machine.
Not equality of wealth.
Not equality of outcome.
Equality of scrutiny.
The billionaire and the builder. The minister and the mechanic. The multinational and the corner shop.
All appearing inside the same analytical system.
AI would not achieve this because it understands justice.
It may simply be unimpressed by status.
And I, for one, would welcome that.
History may not look kindly on those who learned how to game the system while the average person simply carried on working, paying tax and playing by the rules.
The Risk of the Unaccountable Machine
This is not automatically a utopia.
A government capable of reconstructing almost anyone’s financial and physical life possesses enormous power. False positives, bad data, hallucinated connections and political interference could ruin lives.
As analysis becomes cheaper and faster, the main line of defence for the powerful may shift.
Not from hiding evidence.
From challenging the machine’s conclusions.
Or worse, trying to build exemptions and blind spots into the system itself.
Who controls the system, who can inspect its conclusions, and who writes the rules of exemption may matter more than the technology itself.
A society can become more equal in enforcement while also becoming more authoritarian.
That tension matters.
The central political question may not be whether AI will investigate people.
It almost certainly will.
The real question is who gets investigated, who remains outside the machine, and who holds the keys to the blind spots.
For years, ordinary people have been told that surveillance is harmless if they have nothing to hide.
AI may eventually test that principle properly.
Not only on commuters, benefit claimants and motorists, but on the politicians approving surveillance laws, the executives moving profits across borders, the regulators overlooking inconvenient facts, and the wealthy whose finances have always been too complicated to examine.
The machine may know exactly who they are.
It may know their position, their wealth, their title, their connections and their influence.
It just may not care.
We keep asking which workers should fear artificial intelligence.
Perhaps we should also be watching the people who believe they are above scrutiny.
The great disruption may not arrive when AI learns how to do our jobs.
It may arrive when it learns how to follow the money.
And the people who thought they controlled the machine realise that the machine has started looking back at them.