Newsletters
There are days when you walk into a bank expecting to discuss money and instead discover that prejudice has learned to speak fluent mathematics. It arrives disguised as an algorithm, wrapped in government paperwork, politely informing you that discrimination is now a feature rather than a bug.
The unsettling thing about artificial intelligence is that it rarely invents humanity's flaws. It studies them, quantifies them, and hands them back with the reassuring confidence of a machine. If you want to understand what an algorithm truly believes, examine the civilization that taught it to think.
France is an ideal place to begin. This is, after all, the nation that gave the world the Enlightenment, rationalism, the metric system, and endless lectures about how everyone else ought to organize society. Yet somewhere between Voltaire and the compliance department of a government-owned private bank, reason appears to have been mugged in broad daylight.
The scene recently unfolded inside the private banking subsidiary of France's state-owned postal bank, where executives arrived bearing apologies with remarkable enthusiasm. One manager had already been sacrificed to the gods of institutional accountability, while two relationship managers who had spent months questioning perfectly ordinary transactions from a multinational Turkish company had quietly disappeared into the bureaucratic mist.
The bank acknowledged that the company was entirely legitimate. Its transfers were lawful. They were not considered proceeds of criminal or terrorist activity.
Excellent. Civilization survives another day.
Unfortunately, the bank could not bring itself to put any of that in writing.
The explanation that followed belongs in the Louvre.
Yes, the transfers are legitimate.
No, the bank cannot promise future transfers will not be delayed, reversed, frozen, or otherwise treated as though Vladimir Putin had wired the money from inside a volcano.
Apparently, this is not considered contradictory.
It is simply compliance.
The bank operates under two compliance systems. The first belongs to the bank itself. It reviews documentation, examines transaction history, and approves recurring payments.
The second belongs, ultimately, to the French Ministry of Finance, which oversees the bank on behalf of the French government.
Then came the truly unsettling revelation.
According to a source familiar with the Ministry of Finance's procedures, much of this secondary review relies heavily on artificial intelligence platforms such as NICE Actimize and, in some departments, Oracle Financial Crime and Compliance software, with only limited human oversight.
Suddenly, the entire affair makes a peculiar kind of sense.
The prejudice may no longer belong to people.
It may belong to probability.
Algorithms do not wake up distrusting Turks—or anyone else. They inherit statistical assumptions from historical data, regulatory priorities, and decades of accumulated human suspicion. Feed enough biased history into a machine, and discrimination acquires a dashboard.
The system merely flags those transactions based on zip code and geography because that's what the old data told it to do.
The bank's own compliance team may remember yesterday.
The algorithm begins again today.
Unlike a human bureaucrat, the machine has no memory of trust—only a memory of risk.
The first system remembers.
The second behaves like a goldfish suffering from institutional amnesia.
Every transfer is treated as an entirely new event. Previous approvals vanish. Documents already submitted become documents never seen. The same evidence is requested again and again, as though bureaucratic memory resets after every successful transaction.
One is tempted to christen the second system "Godot." It is always there, everyone insists, but no one can say precisely when it will appear, what it will decide, or why it never remembers the last encounter.
The remarkable part is that everyone involved appears perfectly comfortable living inside this administrative Mobius strip. The bank cannot persuade the government. The government ignores the bank. Customers become Vladimir and Estragon in an endless theatrical production called “Waiting for Compliance.”
Financial transactions originating from Türkiye, it seems, are not granted the luxury of presumed legitimacy upon entering the European Union or Great Britain. They begin life beneath a cloud of suspicion, regardless of who sends them, who receives them, or how many times identical transactions have already been approved.
That is not risk management.
It is institutional prejudice disguised as process.
The irony borders on hallucinatory.
Only months ago, at the fifth "No Money for Terror" conference in Paris, Turkish Finance Minister Mehmet Simsek stood beside his French counterpart, Roland Lescure, before representatives of more than 70 national delegations. The gathering celebrated international cooperation against terrorist financing, illicit financial flows, and financial crime, with France publicly praising Türkiye's commitment and effectiveness.
Meanwhile, inside one of the French government's own financial institutions, legitimate Turkish transactions continued entering the compliance system carrying the digital equivalent of a scarlet letter.
Diplomacy applauds.
Compliance suspects.
Apparently, the French government's left hand and right hand communicate through separate fax machines.
This is the grand illusion of the artificial intelligence age. We imagine algorithms eliminate bias when they often preserve it with far greater efficiency. Yesterday's bureaucrat needed coffee before stereotyping an entire country. Tomorrow's machine can accomplish the same feat millions of times before breakfast while producing colorful analytics that appear to prove its own impartiality.
The larger story is not one bank account or one multinational Turkish company. It is about a civilization increasingly surrendering judgment to algorithms while pretending those algorithms arrived untouched by culture, politics, fear, or history.
Machines do not invent prejudice. They industrialize it.
When governments mistake algorithmic confidence for objective truth, bureaucracy ceases to be merely inefficient. It becomes something stranger: a society governed by software that inherited humanity's oldest biases while convincing everyone it has transcended them.
Set as preferred source
Set as preferred source
7 min read
