Korean Deepfake: 17,378 Sexual Violence Cases in 8 Months

Picture a 16-year-old girl who opens a group chat and finds her own face on a body that is not hers. She never posed for it. She never sent it. Everyone in the chat has already seen it. That is the everyday shape of the Korean deepfake problem, and the newest count is hard to shrug off: 17,378.

Korean deepfake abuse: 17,378 reports and a harm that is not hypothetical

Here is the number again, with the context that makes it land. According to Korea JoongAng Daily, the regulator reviewed 17,378 sexual deepfake cases from January through August 2026. All of 2025 came to 15,808. So eight months of this year already beat the whole of last year, by 9.9%.

You might think we have heard this before. Maybe. But a count like this is different from a scary prediction. It is a tally of things that already happened to real people in one country, in less than a year. If you have been telling yourself this is a "future AI problem," 17,378 is the number that should end that thought.

Why generated images still cause real harm to victims

People sometimes say, "But it's fake, so who is hurt?" Ask a victim. A generated image (a picture an AI program made from scratch or edited from a real photo) does not need to be believed to do damage. It only needs to be seen. The embarrassment, the whispers, the lost trust at school or work all land on the real person whose face was used. Researchers describe the fallout for victims as job loss, damaged careers, broken relationships, and in the worst cases suicide.

Who is on the receiving end of Korean deepfake violence against women?

Mostly women. The research behind this story says 96.6% of victims in 2024 were women. This is not random. It is gender based violence that happens to use new technology, and the Korean data shows it plainly. The same research notes that nearly 62% of suspects in one recent year were teenagers. So we are not talking about shadowy hackers. We are talking about kids with cheap apps, picking on classmates, teachers and friends.

---

South Korea deepfakes: the trend, the teens and the schools

The yearly climb is the part that should worry you most. Reports to the regulator went from 1,913 in 2021 to 23,109 in 2024. They dipped to 15,808 in 2025, and some people hoped the worst was over. It was not. The 2026 pace says the dip was a pause, not a cure. This article is part of a series, start with Deepfake Voice Scam News Today 77 Of Targets Lost Money.

Cheap AI tools fuel teen-driven rise in deepfake sex crimes in South Korea

Headline of a report in Korea Herald

That headline says a lot in one line. The tools are cheap, and the people using them are young. A teenager does not need training or money. A phone and a classmate's profile picture will do.

Sexual deepfake pornography moves through schools

Schools are where sexual deepfake pornography (fake explicit images built from a real person's face) hits hardest, because everyone knows everyone. South Korea's Ministry of Education estimated 920 school-based victims of AI-generated intimate imagery. One advocacy center counseled 86 underage victims in 2023. In just the first eight months of 2024, that jumped to 238. Those are kids who were counted, which usually means many more were not.

Think about what that does to a 14-year-old. It follows her from the hallway to the group chat to her phone at night. Nobody has to touch her for it to feel like an attack.

What many people assume	What the Korean numbers show
Deepfake abuse targets famous people	Victims are mostly ordinary women and girls, with 96.6% of victims in 2024 being women
Offenders are organized criminals	Nearly 62% of suspects in one recent year were teenagers using cheap apps
Strict laws make the problem go away	Reports still rose 9.9% after Korea's 2024 law changes
The content lives on big, reachable platforms	17,116 of the 17,378 cases came from offshore pornography sites and peer to peer networks (where users swap files directly)

---

Is a Korean deepfake sexual violence? What the law says and where it falls short

Yes, and Korea treats it that way. Making or sharing a sexual deepfake of a real person is a form of sexual violence, even when no one was touched, because it attacks the person's dignity, safety and control over their own image. After legislation in September 2024, producing or distributing this material can bring up to seven years in prison. Even possessing or viewing it can bring up to three.

Korean deepfake law: Korea's tough rules versus a borderless internet

So why did reports keep climbing? Because the law stops at the border. Of the 17,378 cases, 17,116 came from offshore sites and peer to peer networks. Police in Seoul can arrest a teenager in a classroom. They cannot easily shut down a server in another country or unlock an encrypted chat. This is the enforcement gap, and it is the hardest part of the story. A strong law is a good start. It is not a delete button.

A paper on arXiv puts the problem in its title: "Deepfake Pornography is Resilient to Regulatory and Platform Shocks." Squeeze one site and the content pops up somewhere else. That is also why this is a human rights question, not just a tech one. The rights to privacy and safety of millions of women are being tested faster than any court can react. Previously in this series: Age Verification Adults Flagged As Kids Get Pushed For Id.

---

What can ordinary people do about a Korean deepfake or any sexual fake?

Start with the habit that costs nothing. Treat any unexpected explicit image, or any accusation that comes with one, as unverified until you know where it came from. Do not forward it "just to ask if it's real." Every forward spreads the abuse a little further. If it shows someone you know, tell them privately and calmly before anyone else.

Here is the one thing I would do today. Pick the two or three people who would be the first to see a fake of you, a partner, a parent, a close friend, and tell them now: "If you ever get something like that from me or about me, call me before you believe it." It takes thirty seconds. And it works because the information that shuts a fake down fastest is a person who already knows to doubt it.

If you have ever wondered whether a photo or profile is really who it claims to be, that is the exact question facial comparison technology exists to answer. It lets a trained person line up two images and see whether they point to the same face or whether something is off. At CaraComp, that is the kind of careful, side by side checking we care about. It will not fix the offshore problem. But it is a way to move from "this feels fake" to "here is why."

If a fake of you does turn up, save the evidence first: screenshots, links, dates, usernames. Then report it to the platform and to the police. Do not delete the messages before you have copies. Korean victims are often told to document everything, and that advice travels well.

One last thing to chew on. Of those 17,378 cases, all but 262 lived on sites outside the reach of the usual rules. So when a fake of your face is out there, the question may not be "Is it illegal?" It may be "Will the people who love you stop and doubt it before they hit send?" Up next: Deepfake Voice Scam News Today 77 Of Targets Lost Money.

Korean deepfake: Frequently Asked Questions