The Bot That Undressed a Hundred Thousand
A 2020 investigation by the visual-intelligence firm Sensity found that a bot on Telegram had generated fake nude images of more than 100,000 women from ordinary photographs. Most users who created the images said they were targeting women they knew, and a limited number of the pictures appeared to depict minors.
In October 2020, the deepfake-detection firm Sensity, formerly known as Deeptrace, published an investigation into an ecosystem of automated tools operating on the messaging app Telegram. The report, titled “Automating Image Abuse,” described a bot that let anyone submit an ordinary clothed photograph of a woman and receive back a fabricated nude image, generated within minutes and at no cost.
As of the end of July 2020, at least 100,000 women had been targeted by the bot, MIT Technology Review reported; Forbes, citing Sensity's own count, put the figure at 104,852. The bot required only a single photograph to operate. “If you have ever shared publicly one photo on social media, you may be under threat,” Giorgio Patrini, Sensity's chief executive, told Forbes.
Ordinary women, not celebrities
The images were not primarily of public figures. A poll circulated within the bot's community found that 63 percent of users said they wanted to undress private individuals they knew in real life, against 16 percent who were interested in celebrities and influencers, the AFP news agency reported. The photographs used to feed the bot were typically taken from social media accounts or from private material.
The bot and its affiliated channels had attracted roughly 101,080 members worldwide, with about 70 percent based in Russia and other eastern European countries, according to Sensity's findings as reported by Forbes. The material was created and traded openly across public Telegram channels that could be found through keyword searches.
Images of minors
A limited number of the images shared publicly appeared to depict people who were underage, Sensity found. “Usually it's young girls,” Patrini told MIT Technology Review. “Unfortunately, sometimes it's also quite obvious that some of these people are underage.”
A tool built to be misused
The technology appeared to descend from DeepNude, a 2019 application whose developers took it offline within days of its release, saying the “probability that people will misuse it is too high,” as AFP reported. Sensity warned that the fabricated images could be “explicitly weaponized for the purposes of public shaming or extortion-based attacks.”
Researchers who study image-based abuse note that such fabrications do not require a victim to have done anything at all; a single public photograph is enough. In the years since the Sensity report, several governments have moved to criminalize the creation and sharing of non-consensual sexual deepfakes, though enforcement across borders and platforms remains uneven.