CampusDownWiki
Total citations 125,995한국어KO 12,859EnglishEN 77,485日本語JA 9,775中文ZH 25,876
CampusDown ↗
English

Deepfake

🎨 Media, Arts & Culture By CampusDown Wiki Editorial Team Last updated Views 4
A deepfake is a video, photo or audio recording that seems real but has been made or altered with artificial intelligence, often to show someone saying or doing something they never did.
Contents
  1. 1. What a deepfake is
  2. 2. How deepfakes are made
  3. 3. Why people worry
  4. 4. The law is catching up
  5. 5. Not all bad
  6. 6. FAQ
  7. 7. References

You see a video of a celebrity saying something outrageous. It looks real, it sounds real... but it never happened. Welcome to the deepfake era.

Flat illustration of a computer monitor showing a generic face split in half, natural on one side and a glowing wireframe mesh peeling away on the other, with a magnifying glass on the seam and warning triangles
Half real, half rendered.

What a deepfake is

A deepfake is a video, photo or audio recording that seems real but has been manipulated with artificial intelligence. The name mixes "deep learning," a type of AI, with "fake." The technology can swap faces, change expressions, create faces that don't exist and copy someone's voice.

How deepfakes are made

Comparison chart of deepfakes: how they are made (train a neural network on many images, swap or change a face, synthesize a voice, refine with GANs) versus clues to spot them (odd blinking, mismatched details like earrings, blurry edges, check the source)
How they're made vs. how to catch them.

Most deepfakes use neural networks trained on hundreds or thousands of images of a face. The U.S. Government Accountability Office (GAO) explains two common methods: autoencoders, and GANs, where one network makes a fake and another tries to catch it, over and over, until the fake looks convincing.

Why people worry

GAO notes that deepfakes have mostly been used for nonconsensual pornography that disproportionately targets women, and could also be used to sway elections or spread disinformation. Another risk: once fakes are common, people may dismiss real evidence as "just a deepfake."

The law is catching up

In May 2025 the US enacted the TAKE IT DOWN Act, which makes it a federal crime to knowingly publish nonconsensual intimate images, including AI-made ones, and requires platforms to remove them after a valid request.

Not all bad

The same tools can dub movies into other languages, let shoppers try on clothes virtually, or recreate a voice for someone who lost theirs.

FAQ

Where does the word deepfake come from?
It combines "deep learning" and "fake."
How can I tell if a video is a deepfake?
Look for odd blinking, blurry face edges and mismatched details, and check where the video first came from.
Are deepfakes illegal?
Making one isn't always illegal, but sharing nonconsensual intimate deepfakes is a federal crime in the US under the TAKE IT DOWN Act.

Sources reviewed: 2026-11-20

References

  1. U.S. GAO. Science & Tech Spotlight: Deepfakes (GAO-20-379SP, 2020). Definition, autoencoders and GANs, detection.
  2. The White House. President Signed S. 146 into Law (TAKE IT DOWN Act, May 2025). Federal law on nonconsensual intimate images.

Found this helpful? Share it

Send it to a friend who needs it.

X Facebook Threads

Reports and materials that use this theory (Korean)

There are no materials dedicated to this theory yet, but you can search related reports in the CampusDown library (Korean).

Find more reports · Search '딥페이크' on CampusDown →

You may also like

This article was last updated on November 20, 2026. It is based on widely recognized original works and textbooks; when citing it in a paper, please check the original sources listed in the references.