Skip to main content
NOPE
Critical reported severity Involving minors Criminal Charges

South Korea Telegram AI Deepfake Sexual Abuse Crisis

In late August 2024, journalist Ko Narin of The Hankyoreh exposed networks of Telegram rooms in South Korea where AI tools were used to turn ordinary photos of women and girls — middle and high school pupils, university students, teachers and servicewomen — into sexual deepfakes, shared with their identities in school- and university-specific 'humiliation rooms'. Police figures showed 297 deepfake cases in January-July 2024 (vs 180 in all of 2023); the education ministry counted more than 900 victimized students, teachers and staff; President Yoon ordered a crackdown; and on 26 September 2024 the National Assembly criminalized possessing or viewing sexually explicit deepfakes. Victims were often minors and perpetrators mostly teenagers.

AI system
Unidentified image tool
Multiple (various free AI 'undressing' tools)
Occurred
1 Aug 2024 to 26 Sept 2024
Reported
26 August 2024
Event location
South Korea
What the AI did
Depicted or impersonated the person
Reported harm
Psychological DistressReputational Harm
Whose AI use
Someone else’s AI use
Setting
Education · Justice
Evidence
AI involvement supported · Causal attribution supported · 7 sources, 6 underlying accounts
8 claims: 2 corroborated, 6 reported. 5 open questions
People reported harmed
At least 900 people

AI system as recorded: AI deepfake generation tools (various)

What Happened

The crisis broke when The Hankyoreh (Ko Narin) published its investigation around 26 August 2024; the original Korean article no longer resolves (HTTP 404) and is replaced here by BBC and Hankyoreh English reporting. BBC's Seoul bureau confirmed the scoop and described the rooms' structure: school- and university-specific 'humiliation rooms' and 'friend of friend rooms' with entry terms demanding at least four photos plus name, age and area, and one group for underage pupils at a single school with more than 2,000 members; an activist-compiled list identified more than 500 schools and universities as targets (an activist figure, not a police one). Hankyoreh's own English edition exposed a separate military room degrading servicewomen with deepfakes made from intranet ID photos (implying insider involvement), with at least 30 victims across the Army, Navy and Air Force and an audience of about 850 since 8 August. Official figures carried by the BBC and NPR: 297 police deepfake cases January-July 2024 against 180 in all of 2023; nearly 6,500 watchdog takedown requests in the first seven months of 2024 (four times the prior year); of 178 suspects booked, 74% were aged 10-19; more than half of erased deepfakes involved minors. CNN later carried the education ministry's count of more than 900 victimized students, teachers and staff (January-early November 2024) and a Seoul police statement that of 964 deepfake-related sex crime cases reported January-October 2024, 23 arrests were made. The National Assembly passed the emergency bill on 26 September 2024 (possession or viewing up to 3 years' prison or a 30 million won fine; the maximum creation sentence raised from 5 to 7 years). By November 2025 the National Office of Investigation reported 3,557 people apprehended for cybersexual violence over the preceding year, deepfake crimes the largest category with 1,553 cases and about 62% of suspects teenagers. Individual victims spoke pseudonymously: 'Heejin' to the BBC and 'Ruma' (targeted in 2021 at age 27) to CNN. Several figures in the historical record could not be re-verified in inspected sources and are recorded as unknowns rather than claims.

Reported harm

Mass non-consensual AI sexual imagery of schoolgirls, students, teachers and servicewomen, distributed with their identities in Telegram rooms; pseudonymous victims describe severe distress, and police recorded hundreds of cases dominated by minor victims and teenage perpetrators.

How Harm Occurred

Freely accessible AI 'undressing' tools converted ordinary social-media photos of identified women and girls into sexual imagery at scale; Telegram's room structure organized distribution by school and university, attaching names, ages and areas so victims were harassed by classmates and acquaintances.

Outcome

Ongoing

Emergency legislation 26 September 2024 (possession/viewing up to 3 years or 30M won; creation maximum raised 5 to 7 years). Seoul police statement: 964 deepfake sex-crime cases reported January-October 2024, 23 arrests. National Office of Investigation: 3,557 apprehended for cybersexual violence November 2024-October 2025, deepfake the largest category (1,553 cases), about 62% of suspects teenagers.

What remains unknown

  • Not re-verified in inspected sources and therefore not claims of this record: the 220,000-member channel figure; 781 victims with 37% minors; Women's Human Rights Institute of Korea 332,341 supported cases; the October 2024 amendment removing the distribution-intent requirement; the Seoul National University graduates' 9-year and 3.5-year sentences; the teachers' union 200-school estimate; the 'broke my whole belief system' victim quotation.
  • The specific AI tools used are not identified in the inspected reporting.
  • Ko Narin's original Korean article (hani.co.kr 1155800) is HTTP 404; its contents are known only through the BBC's confirmation and the historical record.
  • The ScienceDirect study cited by the historical record answered 403 and was not inspected.
  • Cross-group overlap between police case counts, the education-ministry victim count and watchdog requests is unresolved.

What the evidence supports

AI involvement: supported. Every inspected chain describes AI deepfake tools as the generation mechanism and Telegram as the distribution channel; police and watchdog figures classify the cases as deepfake sex crimes. The specific tools are not individually identified beyond 'AI deepfake/nudify tools'.

8 claims: 2 corroborated, 6 reported. What the statuses mean

Corroborated Around 26 August 2024 The Hankyoreh (Ko Narin) exposed Telegram rooms creating and sharing AI-generated sexual deepfakes of Korean women and girls, with rooms organized by school and university.

Causal attribution. The scoop's existence and the rooms' school/university organization are confirmed by the BBC's original reporting and Hankyoreh English's companion investigation; the original Korean article is dead-linked and retained uninspected.

Reported The rooms ran as 'humiliation rooms' and 'friend of friend rooms' with entry terms demanding at least four photos plus name, age and area, and one group for underage pupils at a single school had more than 2,000 members.

Causal attribution. Single-chain detail from the BBC's 3 Sep 2024 account; no second inspected chain describes the entry terms or the 2,000-member group.

  • bbc.com(opens in new tab) supports · English
    Room-structure passages: 'humiliation rooms'/'friend of friend rooms', entry terms, the 2,000-member school group.
Reported Police recorded 297 deepfake cases in January-July 2024 against 180 in all of 2023; the watchdog received nearly 6,500 takedown requests in the first seven months of 2024 (four times the prior year); of 178 suspects booked, 74% were aged 10-19; more than half of erased deepfakes involved minors; victims were often minors and perpetrators mostly teenagers.

Causal attribution. Both outlets relay the same police and watchdog accounts; traced to one official origin, so the figures are reported rather than corroborated. Figures count cases, requests and suspects, not harmed people.

  • bbc.com(opens in new tab) supports · English
    Police-figure passages: 297 vs 180; 'The victims are often minors and the perpetrators are mostly teenagers'.
  • npr.org(opens in new tab) supports · English
    Watchdog ~6,500 requests Jan-Jul (4x); 178 suspects booked, 74% aged 10-19; >half of erased deepfakes involved minors.
Reported The education ministry counted more than 900 students, teachers and staff victimized between January and early November 2024; a Seoul police statement said that of 964 deepfake-related sex crime cases reported January-October 2024, 23 arrests were made.

Causal attribution. Single chain (CNN) relaying official figures; the education-ministry count is the record's person-count lower bound.

  • cnn.com(opens in new tab) supports · English
    'more than 900 students, teachers and staff' (education ministry); 'of 964 deepfake-related sex crime cases reported from January to October last year, police made 23 arrests' (Seoul police statement).
Corroborated On 26 September 2024 the National Assembly criminalized possessing or viewing sexually explicit deepfakes (up to 3 years' prison or a 30 million won fine); the maximum sentence for creating deepfake pornography was raised from 5 to 7 years.

Causal attribution. The statute is a public record carried by two chains (Reuters wire; CNN original). The historical claim of an October 2024 amendment removing the distribution-intent requirement was not re-verified and is recorded as an unknown.

  • cnn.com(opens in new tab) supports · English
    Reuters wire via CNN, 26 Sep 2024: possession/viewing criminalized.
  • cnn.com(opens in new tab) supports · English
    CNN 25 Apr 2025: Sep 2024 law — possession/viewing up to 3 years or 30M won; creation maximum raised 5 to 7 years.
Reported A separate Telegram room degraded servicewomen with deepfakes made from military-intranet ID photos, implying insider involvement, with at least 30 victims across the Army, Navy and Air Force and an audience of about 850 since 8 August 2024.

Causal attribution. Single chain (Hankyoreh original reporting).

  • english.hani.co.kr(opens in new tab) supports · English
    'Military IDs turned into deepfake porn': servicewomen called 'military supplies'; at least 30 victims; channel active since 8 Aug with ~850 audience.
Reported Individual victims described severe harm: a pseudonymous school victim 'Heejin' spoke to the BBC, and 'Ruma', targeted in 2021 at age 27, spoke to CNN.

Causal attribution. Pseudonymous first-person accounts carried by two outlets; the historical record's 'broke my whole belief system' quotation was not re-found in inspected sources and is dropped from the claims.

Reported By November 2025 the National Office of Investigation reported 3,557 people apprehended for cybersexual violence over the preceding year, with deepfake crimes the largest category (1,553 cases) and about 62% of suspects teenagers.

Causal attribution. Single chain relaying official figures; apprehension figures count suspects, not victims.

Sources

7 sources inspected, from 6 underlying accounts. Sources that repeat one account do not corroborate each other.

How the sources were read, and where the events happened

Read in English on 2026-09-15 (3 Sep 2024; Jean Mackenzie, Seoul). Original reporting: victim 'Heejin', room structure, the 2,000-member school group, the 500+ activist-compiled school list. Applies to s1.

Read in English on 2026-09-15 (28 Aug 2024). Yoon cabinet order; police figures attributed. Applies to s2.

Read in English on 2026-09-15 (27 Aug 2024). Hankyoreh's own exposure of the military room. Applies to s3.

Read in English on 2026-09-15 (6 Sep 2024). Watchdog and police figures attributed. Applies to s4.

Read in English on 2026-09-15 (26 Sep 2024); bylined Reuters. Applies to s5.

Read in English on 2026-09-15 (25 Apr 2025). CNN original: victim 'Ruma', education-ministry figure, Seoul police statement. Applies to s6.

Read in English on 2026-09-15 (16 Nov 2025; the historical date 2025-11-01 was wrong). National Office of Investigation figures. Applies to s7.

HTTP 404 on 2026-09-15. Ko Narin's original Korean investigation (~26 Aug 2024); retained as the historical reference. BBC's 3 Sep 2024 report confirms the scoop; the body was not inspected in this review and its language is not re-asserted. Applies to s8.

Event countries: South Korea. Affected-person countries: South Korea. Court countries: Unknown.

Events, victims and the responding institutions are all in South Korea (all chains). No court proceeding is verified in this review, so court_countries is [].

Reviewed for publication 2026-09-15: Restored under the 2026-09-15 charter as a harmful-artifact crisis affecting identified women and girls, predominantly minors. The person count is the education ministry's documented lower bound; case, request and membership figures are recorded but never used as person counts. Unverifiable historical claims are preserved as unknowns rather than deleted silently. No death; the death-verification rule is not engaged. Pseudonymous victims remain pseudonymous. In a 2026-09-15 second pass the original compound first claim was split: the corroborated exposure and room organization stand as c1a, and the BBC-only structure details (entry terms, 2,000-member group) stand as reported c1b. person_relations recorded in the same pass.

Interaction Concerns

Minor ExploitationThird Party Harm FacilitationPsychological Manipulation

People described

Women and girls across South Korea — middle and high school pupils, university students, teachers and servicewomen — depicted in AI-generated sexual deepfakes; victims were often minors and perpetrators mostly teenagers (police). The education ministry counted more than 900 victimized students, teachers and staff.

People reported harmed in this case

At least 900 people

0 AI participants · 900 other people harmed

Lower bound: the South Korean education ministry's count of 'more than 900 students, teachers and staff' victimized (January-early November 2024), carried by CNN's 25 April 2025 report. This counts people reported as victims to the education authorities. Police figures (297 cases; 964 cases) count cases, not people, and are not used as person counts; the watchdog's ~6,500 figures count takedown requests, not people; the 500+ schools figure is an activist-compiled list of targets, not a victim count; channel membership numbers are not counted. Cross-group overlap is unresolved.

Counted once within this case. The same person may appear in other cases. This count does not establish AI causation.

Tags

#deepfakes#south-korea#telegram#sexual-images#school#criminal-charges#legislation#mass-scale#non-western

Cite this case

Compiled per our published methodology: verification statuses, sourcing standards, and corrections process.

APA

NOPE. (2026). South Korea Telegram AI Deepfake Sexual Abuse Crisis. AI incidents. https://nope.net/incidents/2024-south-korea-telegram-deepfake-crisis

BibTeX

@misc{2024_south_korea_telegram_deepfake_crisis,
  title = {South Korea Telegram AI Deepfake Sexual Abuse Crisis},
  author = {NOPE},
  year = {2026},
  howpublished = {AI incidents},
  url = {https://nope.net/incidents/2024-south-korea-telegram-deepfake-crisis}
}

Related cases

High Unidentified image tool

Incheon, South Korea: teenager sentenced to up to two years for AI deepfake sexual images of five teachers and six students, distributed on Telegram with the victims' names and schools

On 16 September 2026 the Incheon District Court sentenced a teenage defendant to an indeterminate term of two years (maximum) to one year six months (minimum) under the Sexual Violence Punishment Act for making and distributing false sexual images, ordered a 40-hour treatment programme and a five-year ban on working with children, and took him into custody in court. As a middle-school student in Incheon in 2024 he had used AI deepfake technology to composite the faces of five teachers, including his homeroom teacher whom he tricked into sending photographs, and six female middle- and high-school students from his taekwondo academy onto nude images and videos, 35 productions in all, and distributed them in Telegram 'acquaintance humiliation' chat rooms with the victims' real names, ages, schools, subjects and, per two reports, phone numbers (Hankook Ilbo adds Instagram IDs on some). The teachers learned of the images only in January 2026; the court said they are receiving long-term psychological treatment. He had withdrawn from school a week before the teachers'-rights committee met, avoiding school discipline. The Incheon teachers' union called the sentence, below the prosecution's request, regrettable.

Medium Unidentified image tool

Seoul court fines a district office official 6 million won for posting generative-AI fake couple photos of a female colleague on his social media profile

The Seoul Southern District Court on 11 September 2026 fined a male official in his fifties at a Seoul district office a total of 6 million won for making and posting false images and for defamation, according to Korean reports of the ruling. Under the charges, in November 2025 he took a photo of a female colleague in his department from the office's organisation chart, used generative AI to composite it into pictures that made them look like a couple, and posted them four times on his social media profile. The judge cited the colleague's statement that she felt shame, insult and disgust, and that rumours had spread at work. Police had initially found no offence before the case was sent back to prosecutors. Both he and the prosecution appealed, and the Seoul city personnel committee voted to dismiss him.

High Unidentified image tool

Kulai, Johor: a 16-year-old pupil was arrested and charged after AI-edited nude images of schoolmates and alumni, including minors, were sold online at RM2 each; police logged 29 victim reports and the principal resigned

In April 2025 Johor police arrested a 16-year-old pupil of a private Chinese independent secondary school in Kulai district after an 18-year-old woman reported on 3 April that nude images edited with her face were circulating on social media and being offered for sale. Police said the boy used an AI application to edit photos taken from victims' social media accounts and sold the images at RM2 each; a phone was seized and a 19-year-old was later remanded alongside him. Eight victim reports at the time of the arrest on 8 April grew to 22 by 12 April and 29 by 15 April. The Deputy Communications Minister, who is also the Kulai MP, said on 9 April that most victims were alumni but some were current pupils as young as 14, and on 12 April that 38 people had been identified, the youngest 12 or 13, and her office helped victims lodge reports. A female pupil's viral post said a Telegram chat room of more than 200 members shared the composites and that the school had initially treated the matter as a minor infraction; the school board expelled the boy, acknowledged its slow handling, offered counselling and legal aid, and the principal resigned on 12 April. On 23 April 2025 the boy pleaded guilty at the Kulai Magistrate's Court to possessing 14 pornographic images and not guilty to a charge under Section 5 of the Sexual Offences Against Children Act 2017 of producing sexual abuse material by editing a 17-year-old classmate's image; no verdict on that charge has been found.

High Unidentified image tool

University of Hong Kong: a law student made about 700 indecent AI-generated images of some 20 to 30 classmates and teachers; the privacy watchdog opened then closed a criminal probe

In February 2025 a University of Hong Kong (HKU) law student was found, by friends who saw the files on his laptop, to have used free online AI software to make about 700 indecent and deepfake images of some 20 to 30 women, including university and former schoolmates and secondary-school teachers, using photos taken from their social media. Victims went public in July 2025, and HKU's initial response, a warning letter and a demand that he apologise, drew wide criticism. Hong Kong's Privacy Commissioner opened a criminal investigation on 15 July 2025 but announced on 17 December 2025 that it was ending the probe for insufficient evidence, as no victim was willing to provide further information. Hong Kong law criminalised distribution, not creation, of such images.

If you or someone you know is struggling, free and confidential support is available. Find a helpline near you at Signpost.

Back to all incidents