AI, deepfakes, and a brutal question: what happens to democracy when no one knows what to believe anymore?
“The most effective deception does not consist of hiding reality. It consists of constructing another one and getting millions of people to live inside it…”
NOTE
ANATOMY OF DECEPTION V — THE CENTURY OF MANUFACTURED LIES concludes the Anatomy of Deception series, an exploration of the new forms of manipulation, perception, and construction of truth in the twenty-first century.
It was preceded by
ANATOMY OF DECEPTION I — THE LIE AS AN INSTRUMENT OF STATE
ANATOMY OF DECEPTION II — THE FACTORY OF TRUTH
ANATOMY OF DECEPTION III — THE FIVE LIONS AND THEIR NARRATIVES
ANATOMY OF DECEPTION IV — WHEN THE VICTIM BECOMES THE CULPRIT
WHEN SEEING NO LONGER MEANS BELIEVING
For centuries, a photograph, a recording, or a voice constituted some form of evidence. They were not infallible proof, but they carried particular weight because they seemed to capture a fragment of reality. An image could demonstrate that someone had been there. A recording could preserve a statement. A voice could identify a person. Technology gradually transformed these records into mechanical witnesses to events, imperfect but sufficiently reliable to influence courts, elections, wars, journalistic investigations, and political decisions.
Artificial intelligence is beginning to destroy that historical certainty. Seeing is no longer enough to believe, because virtually any face can speak, any voice can say something it never said, and any event can be reconstructed with an extraordinarily convincing appearance. A person can appear to deliver a speech that never existed, a leader can be shown ordering an action that was never ordered, and a crowd can appear to have gathered in a place where it never was.
For a long timé, humanity learned to distrust rumors because it understood that words could lie. Later, it had to learn to partially distrust photographs because they could be cropped, altered, or taken out of context. Now a far more complex stage is beginning: humanity will also have to learn to distrust what has movement, sound, gestures, breathing, and the appearance of reality.
The change is far more profound than it appears. We are not merely facing new tools of manipulation. We are facing a transformation in the very meaning of evidence. A photograph no longer guarantees that something happened. A voice no longer guarantees that someone spoke. A video no longer guarantees that a scene existed. And when the instruments that for decades served to prove facts begin to lose their authority, society enters unknown territory.
The problem is not merely technological. It is political, legal, journalistic and, above all, democratic. A society can debate different interpretations of the same event. What is far more dangerous is for it to cease sharing the basic certainty that the event actually occurred.
“For centuries, we argued about the meaning of images. Now we are beginning to argue about whether the images ever existed…”
THE DEEPFAKE: THE LIE THAT LEARNED TO HAVE A FACE
In the past, propaganda needed to persuade through words, carefully selected photographs, repeated speeches, or images taken out of context. Now it can directly manufacture the witness, the president, the soldier, the businessman, or the citizen. A face can utter words it never spoke. A voice can announce decisions that were never made. A video can show a meeting that never happened, a false threat, an invented confession, or a military order that no one ever gave.
That completely changes the scale of deception. The lie no longer needs only a narrative: it can acquire a face, a voice, gestures, silences, movements, and the appearance of authenticity. And the more natural these imitations become, the more difficult it is for the ordinary citizen to distinguish between documentation and fabrication.
The problem is compounded by speed. A fake video can travel around the planet in minutes, appear on millions of phones, provoke fear, outrage, or enthusiasm, and begin influencing political decisions before journalists, specialists, or institutions can demonstrate that it never happened. By the time the correction finally arrives, the first impression may already have taken hold.
The deepfake thus introduces a new form of power: not merely interpreting reality, but visually manufacturing it and putting it into circulation as though it had actually happened.
“Twentieth-century propaganda needed to convince you of a lie. Twenty-first-century propaganda can show you the lie happening before your very eyes…”
SPEED: WHEN THE LIE ARRIVES FIRST
Democracy has an additional problem: verification takes time; lying can take seconds. A false image, a manipulated recording, or an invented statement can appear simultaneously on millions of phones before journalists, experts, institutions, or platforms manage to determine what actually happened. By then, emotion may already have done much of the work.
Speed favors deception because a lie does not need to prove anything before it begins to circulate. It only needs to seem possible, confirm a prejudice, feed a fear, or provoke outrage. Truth, by contrast, requires checking sources, reviewing images, comparing accounts, consulting documents, and reconstructing a sequence of events. While one side investigates, the other is already sharing.
During an election campaign, a fabrication released only hours before voting can be especially dangerous. The same can happen during a war, a financial crisis, a protest, or a terrorist attack. A false image attributed to a government, a candidate, or a military force can alter perceptions when no one has yet had enough time to verify it.
And even when a correction subsequently appears, it does not necessarily reach those who received the original message. The lie may have millions of viewers; the correction, only a fraction. Moreover, the first impression often remains even after it has been disproved.
“Truth can win the investigation and still lose the first twenty-four hours…”
ELECTIONS IN THE AGE OF SYNTHETIC REALITY
A candidate can appear to say something he or she never said. A war can appear to have begun before it actually begins. Electoral fraud can be “proven” through images that never existed. A crowd can be created, enlarged, or erased. The problem is no longer merely the manipulation of opinions: there is now the possibility of manipulating the very raw material from which citizens form their opinions.
During an election campaign, a few hours can alter an election. A fake video released the night before a vote can show a candidate insulting a social group, accepting money, confessing to a crime, or uttering a phrase designed to destroy credibility. Even if the recording is later proven false, the election may already be over before the truth manages to catch up with it.
Synthetic reality also introduces a dangerous paradox: the more accustomed societies become to fabrications, the easier it will also become to deny authentic material. A candidate confronted with a genuine recording will be able simply to claim that it was manufactured by artificial intelligence. In this way, the deepfake not only creates false evidence; it can also weaken the value of genuine evidence.
Democracy requires freedom of opinion, but that freedom presupposes some common foundation of facts. If every political sector can manufacture its own video, its own evidence, and ultimately its own reality, an election risks ceasing to be a competition between projects and becoming a competition between incompatible universes.
“An election ceases to be completely free when citizens must first decide whether what they are seeing ever happened…”
SOCIAL MEDIA: THE FACTORY WITHOUT A DIRECTOR
Old propaganda generally had an identifiable source. A government signed a statement, a newspaper published a front page, a political party delivered a speech, or a broadcaster repeated a slogan. New propaganda can originate with governments, parties, companies, ideological groups, private agencies, anonymous individuals, or automated systems capable of producing and distributing content on a scale that was impossible only a few years ago.
Thousands of accounts can repeat the same version until sheer volume creates the appearance of consensus. A claim that initially has no evidence can begin to seem true simply because it appears again and again on phones, in videos, comments, and posts. Repetition slowly replaces proof, and popularity begins to be confused with truth.
Social media adds another decisive element: algorithms do not need to share an ideology in order to amplify a falsehood. Often, it is enough for a post to provoke fear, outrage, surprise, or anger for it to generate engagement and be distributed more widely. Thus, a particularly effective lie can find in the platform’s own design an extraordinarily fast transportation system.
And perhaps therein lies one of the most disturbing characteristics of the new deception. There is no longer always a visible director coordinating the entire operation. Millions of users, algorithms, political interests, economic incentives, and automated systems can participate simultaneously in manufacturing a collective perception without anyone completely controlling the final outcome.
“Old propaganda needed a powerful voice. The new kind may need only millions of echoes…”
THE SECOND DECEPTION: MAKING SURE NO ONE BELIEVES ANYTHING
Here, perhaps, lies the most disturbing consequence of this entire transformation. The danger of deepfakes does not consist merely in millions of people believing something false. There is another, even greater danger: that they may begin to distrust even what is true. An authentic recording may be dismissed as artificial intelligence. A real photograph may be declared manipulated. A documented atrocity may be presented as a fabrication. And a genuine voice may become trapped in the same territory of suspicion as an invented one.
The technology that makes it possible to manufacture false evidence also creates the perfect alibi for denying genuine evidence. A leader caught in a compromising conversation can claim that the audio was artificially generated. A military confronted with disturbing images can argue that they were digitally constructed. A company, a political party, or any public figure can attempt to turn technological doubt into a defense.
Thus, the problem ceases to be merely distinguishing truth from falsehood. It becomes something far deeper: proving that truth is still true. The greater the amount of synthetic content in circulation, the greater the possibility that society will develop a kind of epistemological fatigue, a permanent distrust in which everything becomes suspect and no evidence seems sufficient.
That scenario particularly benefits those who already possess power. If everything can be questioned, then uncomfortable facts can also be buried beneath a cloud of competing versions, denials, fabrications, and doubts. It is not necessary to prove that something did not happen. Sometimes it is enough to make sure that no one is completely certain that it did.
The traditional lie attempted to replace one truth with another version. The second deception is more ambitious: it seeks to destroy the very possibility of establishing a common truth.
“When everything can be false, even the truth begins to need to prove that it exists…”
THE NEW ARCHITECTS OF REALITY
For much of the twentieth century, the power to construct narratives was concentrated in governments, major newspapers, television networks, news agencies, and propaganda apparatuses. Today that power has become fragmented, but it has also multiplied. Google, Meta, OpenAI, X, TikTok, other major platforms, governments, intelligence services, news organizations, technology companies, and millions of users participate, in very different ways, in the architecture of what we see, hear, and ultimately consider relevant.
The problem is not to accuse a single company or to assume that there is one secret command behind everything. The problem is systemic. Whoever controls the models, algorithms, distribution, verification, and visibility possesses an extraordinary capacity to influence what millions of people consider real, important, or worthy of attention.
An algorithm can decide what appears first and what disappears after a few seconds. A platform can amplify certain content and reduce the visibility of other content. An artificial intelligence model can produce images, texts, voices, and videos capable of feeding entire information networks. A government can attempt to intervene in that ecosystem; a company can change its rules; an organized group can exploit its weaknesses.
That is why the struggle of the twenty-first century will not be only for control of territories, resources, maritime routes, energy, or markets. It will also be for the ability to organize perception. Whoever manages to decide what appears before our eyes, what is repeated, what is verified, what is considered doubtful, and what disappears from the field of attention will possess a form of power that is difficult to measure with traditional instruments.
The new architects of reality do not need to order millions of people what to think. It may be enough to decide what they will see first, what they will see a hundred times, and what they may never get to see at all.
“The power of the twenty-first century may consist not only in controlling information, but in designing the landscape within which each person believes he or she is looking freely…”
CAN DEMOCRACY SURVIVE WITHOUT A SHARED REALITY?
Democracy does not require everyone to think alike. It exists precisely because we think differently. But it does require a minimum common ground: that an election took place, that a particular person said a particular phrase, that a war began on a certain date, that a figure belongs to the realm of facts and not merely to the realm of preferences.
“A democracy can survive millions of conflicting opinions. What it can hardly survive is millions of incompatible realities…”
“Artificial intelligence does not need to destroy truth in order to become power. It only needs to multiply so many versions that the truth ends up looking like just one more…”
“Perhaps the great political battle of the twenty-first century will not merely be deciding who governs. It will be deciding who still retains the authority to prove that something actually happened…”
“Thus begins a new inequality between truth and falsehood…”
“One needs time to prove itself.”
“The other needs only speed to take hold…”
Author: Mauricio Herrera Kahn
Brief Bibliography
Robert Chesney & Danielle Keats Citron — Deep Fakes: A Looming Challenge for Privacy, Democracy, and National Security, California Law Review, 2019.
Examines how deepfakes can alter public trust, affect democratic processes, and make it possible not only to manufacture false events but also to deny authentic evidence. It is an especially important reference for the concept of the “second deception.”
Claire Wardle & Hossein Derakhshan — Information Disorder: Toward an Interdisciplinary Framework for Research and Policymaking, Council of Europe, 2017.
Examines disinformation as an ecosystem in which false information, manipulation, digital dissemination, and social behavior interact. It is particularly useful for understanding how networks can transform repetition, speed, and virality into the appearance of truth.
Historical and Political Thought
Walter Lippmann — Public Opinion, 1922.
A foundational work on the way people construct their perception of the world through images, stereotypes, and representations that do not necessarily correspond to direct reality. A century later, his reflections acquire a new dimension in the face of algorithms, social media, and artificial intelligence.
Hannah Arendt — Truth and Politics, 1967.
Reflects on the conflict between factual truth and political power. Her thinking is particularly relevant in a world where facts can be manipulated, disputed, or directly denied, and where the loss of a shared reality may ultima
