TOKYO (TR) – Tokyo Metropolitan Police have arrested a 32-year-old company employee for allegedly using generative AI to create and distribute obscene deepfake images of real women online, investigative sources revealed on August 3, reports the Asahi Shimbun (Aug. 3).
Kenta Imoto, a resident of Himeji City, Hyogo Prefecture, was arrested on suspicion of defamation and violating the child pornography prohibition law. Upon his arrest, Imoto admitted to the allegations, telling investigators, “I was happy that [the requesters] were pleased when I processed the images with AI.”
He went on, “I processed about 300 images of girls in their teens and twenties. I focused on track and field uniforms because they were easy to edit.”
Police also plan to send papers to prosecutors regarding a 17-year-old high school boy from Tarumizu City, Kagoshima Prefecture. The teen is suspected of supplying the original photos via social media and requesting Imoto to create the sexually explicit deepfakes.
According to investigative sources, the pair conspired on February 11 to alter a photograph of a woman, now in her 40s, taken during a middle school sports festival when she was wearing gym clothes. Knowing the subject in the original photo was under 18, Imoto allegedly used generative AI to create two nude images of the girl.
The suspects then posted the fabricated images on a social media bulletin board, making them available to an unspecified number of viewers. Police state this act defamed the victim and constituted the public display of child pornography.
Imoto is also suspected of using the same method to create and upload obscene AI-generated images of three other women in their 20s.
Approximately 3,800 obscene images and videos — including child pornography believed to have been created using generative AI — were found stored on Imoto’s smartphone.
Police noted that neither Imoto nor the teenage boy had any personal acquaintance with the victims. Investigators believe the suspects scraped the original photos of the women directly from the internet and social media platforms before manipulating them.
