Serbian women whose AI-generated nude photos have surfaced online are finding themselves defenceless against the emerging technology that is fueling deepfake pornography.
In 2020, Adrijana Petkovic took a seemingly harmless bathroom selfie in her home in Knjazevac, a small town in eastern Serbia. Four years later, the photo surfaced unexpectedly. Her husband received it from a colleague who got it from another coworker.
By this time, the photo had been altered by AI, sexualising Petrovic without her knowledge or consent. This marked the beginning of a harrowing experience for her. "I'm a mother of two young children; I have a family, parents, a stable and peaceful life. This was an enormous shock for us," Petkovic, 25, told BIRN.
As part of their search, Petrovic and her family contacted everyone they could identify who had been involved with sharing the picture. The last trace we found led to a Telegram group called "Knjazevac - Bor" where it appears someone initially posted the image. From there, it spread as people downloaded and forwarded it amongst themselves, Petrovic revealed.
From Eroding Trust To Damaging Reputations
Despite filing a complaint with the police, Petrovic received no updates for two weeks. When she finally heard back, they informed her that they could only issue a warning to the "Knjazevac - Bor" Telegram group administrator.
They explained that since there wasn't blackmail involved, their options regarding sharing the picture within groups and messages were limited. "You're not the first, and you won't be the last," the officer told her.
Petrovic hoped the police could identify the perpetrator who created the deepfake and provide legal recourse. However, the response left her feeling utterly helpless. "I was shocked that it was the most normal thing people do and that you can do absolutely nothing against it. You have no support from the police."
Deepfake pornography is a non-consensual exploitation tactic. It uses artificial intelligence to manipulate someone's image or video into sexually explicit content, primarily targeting women. This investigation revealed that Telegram currently allows access to numerous channels and bots. These tools reportedly leverage AI technology to manipulate photos, creating nude versions in a matter of seconds.
To understand the accessibility of these tools, this investigation involved testing a selection of publicly available bots shared within prominent Balkan Telegram groups. These bots varied in functionality, with some requiring payment and others offering limited free trials. Notably, all displayed a concerning ease of use.
Telegram has been battling a disturbing trend of deepfake nudes targeting women for a while now. In 2020, a cybersecurity report by Sensity revealed that deepfake nude images have been circulating on the messaging app Telegram, primarily targeting under-aged females.
Experts warn that the consequences of sharing these images can be devastating. "The consequences are the same as when a real, unedited sexually explicit photo of a woman or girl is published," said Vanja Macanovic, a lawyer at the Autonomous Women's Centre in Belgrade, which provides legal and mental health support to victims.
Macanovic claims this can be even more distressing for the victim because they know the photo is fake, but they are unable to definitively disprove it. As a result, anyone who sees the image may believe it is real.
In Serbia, where creating these images isn't explicitly illegal, victims have limited options beyond lawsuits. Macanovic emphasises Serbia's need for action, warning the situation will worsen without legal intervention.
"This is how the perpetrators of these crimes in Serbia go unpunished, as well as the platforms through which the crimes are committed," the lawyer told BIRN. Without legislative changes, Macanovic warned, "there will be more and more cases like this, and it will become the dominant form of peer violence against girls as well as violence against women by ex-partners."
Unpunished Abusers Across The Region
The issue of 'deepfake pornography' extends beyond Serbia's borders. In May 2024, Kosovar authorities arrested seven individuals suspected of managing and contributing to the Telegram group "Albkings."
This group, boasting over 100,000 members, allegedly facilitated the sharing of derogatory videos, deepfake images, and personal information.
Neighboring countries are taking action. Croatia introduced a "misuse of sexually explicit content" law in 2021, and Montenegro criminalised the misuse of someone's sexual recordings or images in 2023, with prison sentences of up to two years. In contrast, Serbia lacks any such legislation, leaving victims with limited options.
The ease of manipulating images is deeply troubling. On International Women's Day, March 8, a Telegram group administrator known as "Dzoni" shared manipulated sexual images of popular Serbian singer Goca Trzan. This incident occurred within a group called "Balkan Porn," which boasts over 70,000 members.
The original image, which Trzan shared with her half-million followers, depicted her wearing a grey t-shirt at home and discussing her IVF treatment struggles. "Dzoni" purposefully manipulated the image to exploit Trzan and create a sexually explicit scene.
The manipulative tactics employed by "Dzoni" were swift and disturbing. On the same day, he shared a seemingly innocuous photo in the Telegram group: an unidentified woman with dark hair, wearing a pink summer shirt, sitting in a cafe in the Montenegrin resort town of Budva.
Just moments later, he uploaded a second image – the same woman, in the same location, completely nude. This rapid manipulation highlights the chilling ease with which deepfakes can be weaponised to violate privacy and create explicit content.
It is alarming that young people are getting drawn to this technology's ability to generate nude images. Last month, for instance, police arrested a teenage boy who reportedly circulated AI-generated nude images of 50 high school girls.
While fascinating, deepfakes have a dark side. Their misuse can erode trust in media, sow discord, and inflict immense harm on individual reputations. As AI continues to develop, so must safeguards to prevent exploitation.
