Japan Deepnude Arrests Signal Gen-AI Deepfake Focus for Law Enforcement
Tokyo’s Metropolitan Police Department this week arrested a 31-year-old man for allegedly creating non-consensual deepfake pornography of female celebrities, known as deepnudes, and distributing them on the internet.
The Tokyo Met said suspect Tatsuro Chiba uploaded at least 2,200 deepfake images to paid membership websites since 2023, and made around US$70,000 from subscriptions to view the images.
Investigators say he used free generative-AI software to create the ‘deepnude’ media, generating over 520,000 images of about 300 celebrities, actresses, and pop idols over the space of two years.
The suspect posted samples on promotional social media accounts on X (twitter) and Instagram to attract customers, which were then directed to paywalled sites with a fixed monthly subscription.
Are deepfake crimes a growing trend?
This is one of the first deepnude arrests we have seen, or arrests for any deepfake-related AI crimes, but the trend is quietly growing. In 2025, Tokyo police arrested another 31-year-old suspect, Hiroya Yokoi, who is alleged to have sold around 20,000 sexually explicit deepfake images of 262 women.
While Japan, like most countries, does not have any kind of comprehensive ‘anti-deepfake legislation’, the government is in the process of strengthening its legal framework around deepfakes and AI in general, currently leveraging existing laws like the Penal Code, which bans the distribution of obscene digital content.
But the criminalisation of deepfake pornography, and subsequent deepnude arrests, is a trend that has really only appeared in the last 12 months, in line with the democratization of deepfake generation through ready access to sophisticated and user-friendly generative-AI tools. Trained on even a single source image, these deepfake generators enable criminals to operate at the scale seen above, churning out hundreds of thousands of convincing images, videos, and audio with ease.
Meanwhile, law enforcement and judiciary authorities are still getting to grips with technological developments that are fast outpacing the law. Few countries are leading the charge, but possibly none more so than South Korea, where the National Police Agency (NPA) made 682 arrests over deepfake sex crimes in 2024-2025. However, more than 80% were minors, many of which are exempt from criminal prosecution, highlighting the social challenges of gen-AI tools being made available to everyone, including children and teens. According to the South Korean police, a total of 1,202 deepfake-related sex crime cases were recorded within the same time frame.
Aside from Japan, notable arrests for deepfake pornography offences have taken place in the UK, where the government is cracking down on sexually explicit deepfakes through proposed changes to the law, and 25 arrests over 19 countries by Europol as part of an operation understood to be one of the first cases involving AI-generated child sexual abuse material (CSAM).
What is the impact of deepfake crime on forensics and law enforcement?
Forensic analysts are feeling the pain. It’s their job to do the field work, and investigative units are increasingly confronted with crimes where synthetic media is not incidental, but central, and in these cases, speed matters as much as accuracy. Every hour spent validating manipulated content with outdated deepfake detection tools is an hour lost for victim protection, takedowns, or suspect identification.
Fundamentally, the ease with which malicious media can either be manipulated or completely fabricated is changing the nature of how forensic authenticity is handled in terms of deepfake detection. Given the volumes and sophistication involved it can no longer be treated as a late-stage expert opinion, but must be addressed instead as an automated verification step in the process.
The optimal forensic process should be to integrate AI-generated deepfake detection at the earliest stage of evidence ingestion, so every extracted media file can be automatically screened for signs of manipulation before it enters the investigative record.
By following this approach, analysts report dramatic reductions in manual review time of deepfake detection cases, faster triage of large datasets, and—most critically—greater confidence when introducing digital media into legal proceedings.
