Skip to main content
NOPE
Medium reported severity Criminal Charges

Tokyo police say a man used generative AI to make about 20,000 sexual images resembling 262 female celebrities and sold access to them

The Tokyo Metropolitan Police announced on 16 October 2025 that they had arrested a 31-year-old company employee on suspicion of displaying obscene electronic records, for posting on social media obscene images made with image-generation AI to resemble female celebrities. According to the police, from October 2024 he posted about 20,000 AI-made sexual images resembling 262 actresses, idols and announcers and earned about 1.2 million yen in viewing fees; the charge covers three images resembling three actresses posted between January and June 2025. Police say he trained a freely available image-generation AI on the women's images, let higher-paying viewers name the celebrity to be depicted, and admitted the charge. The police described it as the first arrest over AI-made obscene images resembling celebrities.

AI system
Unidentified image tool
Occurred
Oct 2024
Reported
16 October 2025
Event location
Japan
What the AI did
Depicted or impersonated the person
Reported harm
Exploitation or Abuse
Whose AI use
Someone else’s AI use
Setting
Privacy · Justice
Evidence
AI involvement reported · Causal attribution alleged · 2 sources, 1 underlying account
5 claims: 5 reported. 3 open questions
People reported harmed
At least 3 people

AI system as recorded: A freely available image-generation AI that, according to the Tokyo police, the suspect trained on image data of actresses and idols on his home computer to produce obscene images closely resembling them (tool not identified)

What Happened

The police account. The Tokyo Metropolitan Police announced on 16 October 2025 that they had arrested a company employee, 31, the day before on suspicion of displaying obscene electronic records. According to the announcement as reported by the Yomiuri Shimbun, between January and June 2025 he posted three obscene images resembling three actresses on social media where anyone could view them. Police believe that from October 2024 he posted about 20,000 AI-made sexual images resembling a total of 262 actresses, idols and announcers and earned about 1.2 million yen in viewing fees.

How the images were made. Police say he trained a freely available image-generation AI on his home computer with image data of actresses and idols and produced obscene images closely resembling them. He posted sample images on a paid creators' membership site and gave paying users access to the social-media account where the obscene images were posted, ranking viewers in five tiers; those who paid more could see more explicit images and name the celebrity to be depicted. About 21,500 images of celebrities, believed to have been used to train the AI, were found on devices seized from his home. The case came to light through the police's cyber patrol.

The suspect's statement. Police say he admitted the charge, telling investigators that images modelled on female celebrities drew a large response and were popular, and that he expected to earn a lot from them.

Context. The police said this was the first time they had acted against obscene images resembling celebrities made with image-generation AI. Daily Shincho reported on 25 October that the man was arrested on 15 October by the police's safety division after the cyber patrol. No response from any depicted woman is reported in the sources read.

Reported harm

According to the Tokyo police, a man used image-generation AI to make sexual images closely resembling female celebrities without their consent and sold access to them; the charge concerns three actresses and police believe 262 women were depicted.

Outcome

Unknown

The Tokyo police arrested the suspect on 15 October 2025 on suspicion of displaying obscene electronic records and announced the arrest on 16 October. Police say he admitted the charge. Whether he was indicted or tried is not reported in the sources read.

What remains unknown

  • Which image-generation tool was used.
  • Whether any depicted woman was told of the images or responded.
  • Whether the suspect was indicted or tried, and the outcome.

What the evidence supports

AI involvement: reported. The investigating Tokyo police say the suspect trained a freely available image-generation AI on image data of actresses and idols and used it to produce obscene images closely resembling them, about 20,000 images of 262 women by their estimate, including the three charged images resembling three actresses. Devices seized from his home held about 21,500 celebrity images believed used for the training, and police say he admitted the charge. Those generated sexual depictions of real women are the harm. The tool is not named.

5 claims: 5 reported. What the statuses mean

Reported The Tokyo police announced on 16 October 2025 the arrest of a man on suspicion of displaying obscene electronic records, for posting on social media obscene images made with image-generation AI to resemble female celebrities.

Causal attribution. Police account of the arrest; the AI generation is the police's description.

  • yomiuri.co.jp(opens in new tab) supports · Japanese
    画像生成ＡＩ（人工知能）で女性芸能人らに似せて作ったわいせつ画像をＳＮＳに掲載したとして、警視庁は１６日 ... わいせつ電磁的記録記録媒体陳列容疑で逮捕したと発表した。
  • dailyshincho.jp(opens in new tab) supports · Japanese
    10月15日にわいせつ電磁的記録媒体陳列容疑で逮捕しました。
Reported Police believe that from October 2024 he posted about 20,000 AI-made sexual images resembling 262 actresses, idols and announcers and earned about 1.2 million yen in viewing fees.

Causal attribution. Police estimate as reported; the counts are beliefs of the investigators.

  • yomiuri.co.jp(opens in new tab) supports · Japanese
    昨年１０月以降、女優やアイドル、アナウンサー計２６２人に酷似したＡＩ作成の性的画像約２万点をＳＮＳに掲載し、計約１２０万円の閲覧料を得た
  • dailyshincho.jp(opens in new tab) supports · Japanese
    この262人のわいせつ画像を生成AIで約2万点作り、ネットのコミュニケーションサイト上で公開していた男が警視庁保安課に逮捕された。
Reported The charge covers three obscene images resembling three actresses posted between January and June 2025; police say he admitted it, saying images modelled on female celebrities were popular and he expected to earn a lot.

Causal attribution. Police account of the charge and the suspect's statement.

  • yomiuri.co.jp(opens in new tab) supports · Japanese
    会社員の男は１～６月、女優３人に酷似したわいせつ画像計３点をＳＮＳに掲載し、不特定多数が閲覧できるようにした疑い。調べに「女性芸能人を模した画像は閲覧者からの反響が大きく、人気があった。多くの収益を見込めると思った」と容疑を認めている。
Reported Police say he trained a freely available image-generation AI on image data of actresses and idols to make obscene images closely resembling them, and let higher-paying viewers name the celebrity to be depicted.

Causal attribution. The AI generation of the images, as the police describe it.

  • yomiuri.co.jp(opens in new tab) supports · Japanese
    インターネットに無料で公開されている画像生成ＡＩに女優やアイドルらの画像データを学習させ、本人に酷似したわいせつ画像を作成。 ... 画像のモデルにする芸能人を指定させて、希望通りの画像を作成していたという。
Reported Police say the case was found through their cyber patrol, that about 21,500 celebrity images believed used to train the AI were found on his seized devices, and that it was the first time they had acted against AI-made obscene images resembling celebrities.

Causal attribution. Police account.

  • yomiuri.co.jp(opens in new tab) supports · Japanese
    同庁によると、画像生成ＡＩで作成された芸能人らに似せたわいせつ画像の摘発は初めてだという。 ... 同庁のサイバーパトロールで発覚した。会社員の男の自宅から押収した端末には、ＡＩの学習用に使ったとみられる約２万１５００点の芸能人らの画像データが見つかった。

Sources

2 sources inspected, from 1 underlying account. Sources that repeat one account do not corroborate each other.

How the sources were read, and where the events happened

Read in Japanese by the research agent (an AI) on 2026-10-22 and translated by the research agent (an AI) into English; no human translator or reviewer was involved. Yomiuri Shimbun report of 16 Oct 2025 (datePublished 2025-10-16T12:02+09:00) relaying the Tokyo police announcement; HTTP 200. Applies to s1.

Read in Japanese by the research agent (an AI) on 2026-10-22 and translated by the research agent (an AI) into English; no human translator or reviewer was involved. Daily Shincho report of 25 Oct 2025, HTTP 200; the page shows only the opening paragraphs, which relay the police account, so it shares the announcement's independence group. Applies to s2.

Event countries: Japan. Affected-person countries: Unknown. Court countries: Unknown.

The Tokyo Metropolitan Police investigated and made the arrest, and the images were made on the suspect's home computer in Japan and posted on social media and a membership site, so the event country is Japan. The depicted women are described only as actresses, idols and announcers, so their countries are not recorded. No court proceeding is reported in the sources read.

Reviewed for publication 2026-10-22: Published as a core case (depicted_or_impersonated): the investigating Tokyo police say a man trained a freely available image-generation AI on actresses' and idols' images and made obscene images closely resembling them, three of which, depicting three actresses, are charged; he admitted the charge. Non-consensual sexual depiction is itself the reported harm. No suspect or depicted woman is named.

People described

Female celebrities (actresses, idols and announcers) whose likenesses, according to the police, were reproduced in AI-made sexual images; the charge concerns three actresses, and police believe 262 women were depicted

People reported harmed in this case

At least 3 people

0 AI participants · 3 other people harmed

Additional affected people are described without a reliable count.

Three actresses whose likenesses appear in the three charged images (police announcement as reported by the Yomiuri). Police believe images resembling 262 actresses, idols and announcers were posted; beyond the three charged, they are left unquantified. Viewers who paid fees are not counted.

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

Tags

#sexual_deepfake#generative_ai#arrest#japan

Cite this case

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

APA

NOPE. (2026). Tokyo police say a man used generative AI to make about 20,000 sexual images resembling 262 female celebrities and sold access to them. AI incidents. https://nope.net/incidents/2024-tokyo-police-say-man-used-generative-ai-to-make-sexual-images-resembling-262-female-celebrities-and-sold-access

BibTeX

@misc{2024_tokyo_police_say_man_used_generative_ai_to_make_sexual_images_resembling_262_female_celebrities_and_sold_access,
  title = {Tokyo police say a man used generative AI to make about 20,000 sexual images resembling 262 female celebrities and sold access to them},
  author = {NOPE},
  year = {2026},
  howpublished = {AI incidents},
  url = {https://nope.net/incidents/2024-tokyo-police-say-man-used-generative-ai-to-make-sexual-images-resembling-262-female-celebrities-and-sold-access}
}

Related cases

High Unidentified image tool

Nagoya, Japan: a former elementary-school teacher was sentenced on 4 June 2026 to three and a half years, including for possessing generative-AI nude images made from school photos of two girls; reported as Japan's first conviction applying the child-pornography law to AI deepfakes

On 4 June 2026 the Nagoya District Court sentenced a 35-year-old former Nagoya municipal elementary-school teacher to three years and six months in prison. Prosecutors had sought six years. He was one of seven current and former teachers from five prefectures charged over a group chat in which members shared secretly filmed images of girls. His offences included secret filming and sharing, and, under the child-pornography law, possessing two generative-AI images showing two real girls apparently naked. According to the police and the judgment summary, he sent photos of the girls from the school where he taught to another member of the group, who made the images on an AI image-editing website. The teacher kept them on his phone, and one was shared in the chat. The judge found the images realistic enough to be mistaken for photographs of the girls. Jiji called it the first case in Japan applying the law to possession of AI-generated sexual deepfakes, and Kyodo called it apparently the first such court finding. A lawyer and a former judge argue the law was stretched.

Medium Unidentified image tool

Tokyo police say a man posted over 500,000 AI-made sexual deepfakes resembling about 300 actresses and idols on paid member sites; arrested in January 2026

The Tokyo Metropolitan Police arrested a 31-year-old unemployed man on 19 January 2026 on suspicion of displaying obscene electronic records, for posting on member sites obscene images made with image-generation AI to resemble female celebrities, the Yomiuri Shimbun reported. Police believe that since summer 2023 he posted more than 500,000 sexual deepfakes using images of about 300 actresses and idols and earned about 11 million yen in viewing fees. According to investigative sources, the charge covers a dozen or so obscene images resembling female celebrities posted on two member sites between December 2024 and May 2025. Police say he trained a freely available image-generation AI on the women's images, advertised with sample images on X and Instagram, and took requests from higher-paying members for which celebrity to depict.

Medium Unidentified image tool

Henderson, Kentucky: police identified AI-generated nude images of two middle-school children on a device after a report that such images were being created and shown to other students at South Middle School; a juvenile was lodged at the juvenile detention centre on charges of distributing sexually explicit images without consent and possessing and distributing matter portraying a sexual performance by a minor

Henderson Police said that on 'last Thursday' (17 September 2026) they received a report that AI-generated nude images of children were being created and shown to other students at South Middle School in Henderson, Kentucky. During the investigation officers found and identified AI-generated photographs of two minor victims on a digital device. A juvenile accused of sharing the images was lodged at the Juvenile Detention Center on three charge types: distribution of sexually explicit images without consent; two counts of possession or viewing of matter portraying a sexual performance by a minor over 12 and under 18; and distribution of matter portraying a sexual performance by a minor over 12 and under 18. The school said on Monday 21 September that it was aware of a situation involving a student allegedly misusing artificial intelligence and inappropriate images, that the police investigation was active and that it had remained in communication with the families involved. 14 News (WFIE) reported the case on 21 September; WKYT carried the same report. The tool is not named. The depicted children's exact ages are not reported, though the charge descriptors put them between 12 and 18. The accused is described as a South Middle School student; the accused's age and any relationship to the depicted children beyond attending the same school are not reported.

Medium Unidentified image tool

Jaú, São Paulo: eight 15-year-old ninth-grade boys at a private school are reported to have made fake nude images of a female classmate with artificial intelligence and shared them in a WhatsApp group; the school suspended them and the Civil Police are investigating a 15-year-old for dissemination of child pornography and defamation

In early September 2026 a private school in Jaú, in the interior of São Paulo state, identified the circulation of fake intimate images of a female classmate made with artificial intelligence and contacted the families. According to the police report described by the press, eight ninth-grade boys, all aged 15, kept a WhatsApp group in which they used digital programs to manipulate images of the student so that she appeared unclothed. The school suspended the eight and informed the girl's parents. The father of one of the boys checked his son's phone, found the files and took the boy and the phone to the police to register a report; the girl's parents also registered a police report. The São Paulo Public Security Secretariat said the Civil Police are investigating a 15-year-old for dissemination of child pornography and defamation that occurred on the morning of Thursday 3 September. The school said the content originated on a private platform outside the school environment and that it had adopted protective and welcoming measures for the student and opened an internal inquiry. The AI tool used is not named. The students are minors and are not named in any report.

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