A Society That Demands You Prove What's Real

We think of deepfakes as 'technology that creates fake videos.' That's where the misunderstanding begins.

The real change happens not on the creation side, but on the viewing side. Once generated videos become commonplace, at some point we start viewing all videos with suspicion. In that moment, the center of gravity of proof shifts entirely.

Here's an analogy. In the past, counterfeit bills were the problem. You just had to weed out the occasional fake. Now it's as if every bill comes with a sticker saying "prove this is real." It's not the cost of filtering out fakes, but the cost of proving authenticity that spreads across society as a whole.

In technical terms, you might call it 'transfer of the burden of authenticity verification.' In plain language, it's a structure where not the person spreading falsehoods, but the person who sees it must shoulder the question "is this real?"

Many people stop here. "So do I need to install a deepfake detection app or something?" The moment you narrow the problem down to individual usage, you miss the real point.

Detection can't be solved with individual skills. Even verifying the authenticity of a single video requires time, knowledge, and tools. If 50 million people each bear that burden, society's total verification cost balloons astronomically. Moreover, that cost isn't distributed equally.

That's why we need to redefine trust. Trust is not an attribute embedded in content. It's a question of distribution—who bears the verification cost. A well-designed society absorbs that cost through institutions, not individuals.

In Korea today, this problem manifests most sharply in two places: elections and trials.

Consider elections. Three days before voting, a fake video of a candidate's statement spreads. It takes the National Election Commission days to verify it and request deletion, but by then voting has already ended. The moment verification speed can't keep up with decision-making speed, the fake wins regardless of truth. The Public Official Election Act has provisions on spreading false information, but 'who determines the authenticity of generated videos and by what standards' remains blank.

Trials are even weightier. In court, video has long been powerful evidence. CCTV, dashcams, call recordings—all of these. Now defense attorneys can challenge even real videos with "isn't this generated?" and conversely insist a fake is "real." The very standards for determining evidentiary value are being shaken. Korean courts have yet to establish who bears how much of the burden of proof for generated videos.

There are counterarguments. "Problems created by technology are solved by technology. Just implement watermarks, source authentication (content provenance standards like C2PA), and detection AI." This is only half right. Detection is an eternal arms race of cat-and-mouse with generation, and watermarks can simply be removed. Technical standards are merely the foundation—without institutions above them defining 'who has the authority and responsibility to determine authenticity,' they collapse. Technology only reduces costs; it doesn't distribute responsibility.

This is the institutional gap Korea has left empty. Mandatory source labeling for generated videos, rapid adjudication bodies during election periods, evidentiary standards for courts, and above all the principle that 'the burden of proof lies with platforms and senders, not individuals.' This can't be filled with a single app.

When an elderly person in Busan receives a video of a mayoral candidate in a group chat and asks "is this real?", the responsibility to answer shouldn't rest with that elderly person. It's a public cost that society must design itself to bear.

So the minimum literacy citizens need isn't 'detecting deepfakes well.' It's the sense to recognize who the burden of authenticity verification is currently being shifted onto. Citizens who understand this demand institutional solutions.

We don't need to become a country good at detecting deepfakes. We need to be a country that doesn't shift the responsibility of proving authenticity onto individuals, a country designed so institutions absorb that cost. More than a country good at using AI, a country that properly understands how trust collapses will be the one that survives in the end.