The emergence of artificial intelligence has transformed the world in profound ways, bringing...
Enter the virtual art gallery.
“I don’t think I will be able to get back online, or be healed, or have a normal life as long as these evil-incarnated people are out there with huge internet access and the emergence of AI. I am not sure if I will ever enter another relationship, or build a life and have my own family. I am not sure if I will ever get back my previous, serene life.”
This is the voice of Hawi, 27, a survivor whose life has been severely disrupted by an AI-generated video made using Google’s Veo 3 video-generation tool.
Hawi, whose name has been changed at her request, is living with deep psychological trauma and recurring panic attacks that began after an ex-partner shared deepfake videos of her. Her case illustrates how emerging AI technologies are being weaponized, particularly against women, in contexts where legal and institutional safeguards remain weak.
As a nonprofit journalism organization, we depend on your support to fund more than 170 reporting projects every year on critical global and local issues. Donate any amount today to become a Pulitzer Center Champion and receive exclusive benefits!
A Relationship That Turned into Abuse
Hawi and her former boyfriend were in a relationship for three years. They shared periods of hardship and moments of happiness, struggling together through unemployment and celebrating small successes when things improved.
“I can’t say all three years were smooth. We quarreled and reconciled many times. But we kept moving forward, hoping to get married one day once we were financially stable and psychologically ready,” she recalled.
Over time, however, the conflicts became more frequent and intense. “At some point, I don’t even remember exactly where things went wrong. We started fighting over trivial issues,” she said. The disagreements eventually led to a painful separation. “I once called him and we sat together for several hours talking. I told him firmly that the relationship was over. It wasn’t easy to reach that decision, but the fact that he once attempted physical abuse showed me what our future would look like. That is when I decided to end it.”
Harassment After the Breakup
Following the breakup, Hawi’s situation deteriorated rapidly. For weeks, she received hundreds of threatening and insulting text messages and missed calls from different phone numbers. “I blocked his number the same day we broke up and took time off work to organize myself. Three days later, he called me from another number after realizing I had blocked him. Once I recognized his voice, I told him to stop calling and blocked that number as well.”
Instead of stopping, her former partner escalated the abuse. He began spreading rumors among their mutual friends, claiming that Hawi had left him for a wealthy older man, someone old enough to be her father. According to Hawi, the misinformation spread quickly and had devastating consequences. She withdrew socially, resigned from her job due to workplace gossip, and was questioned by her supervisor about the rumors.
The Deepfake Video
The harassment culminated in the release of an AI-generated deepfake video. “I couldn’t handle it. The video looked so real that it made me question myself. I saw my own image speaking and acting like a happy bride.” The video caused serious turmoil within her family. Hawi said her parents initially believed the content was real.
“One evening, my parents called me to discuss the issue. The entire conversation was about encouraging me to speak openly about the marriage created by AI. They were already convinced it was real and thought I was hiding it,” Hawi said.
That moment pushed Hawi to report the case to the police. She was convinced her former boyfriend was responsible for the AI-generated video. Although he later deleted the video after rigorous mediation by mutual friends, the damage had already been done.
Law Enforcement Response
Hawi said the police response was deeply discouraging. “They barely understood the case. It took a long time just to explain what had happened. When they watched the video, they thought it was real and asked me to bring the so-called husband for questioning.”
She said it took nearly a week for officers to understand that the video was AI-generated.
“Eventually, they told me it would be difficult to hold the suspect accountable because there is no legal framework to address this kind of crime.”
The experience left her disillusioned. “I lost hope. I had already resigned from the job I worked so hard to get. I know the system is not in favor of women. Many outspoken women who advocate for women’s rights or speak publicly about their experiences face retaliation. Some flee the country, others are detained. I don’t believe my fate would be any different.”
Hawi is now trying to rebuild her life. She says she plans to speak openly about her experience and the justice system once she has fully recovered, and is considering advocacy work. For now, however, she has no social media accounts and does not use a smartphone.
The Broader Situation in Ethiopia
AI-generated deepfake videos are becoming more common in Ethiopia. Such content has targeted senior government officials, mayors, regional presidents, the prime minister, and, increasingly, women.
Addisalem Birhane, a senior expert in gender, development and digital inclusion, said many young people are highly technologically savvy and capable of operating clandestinely, including masking their IP addresses to avoid detection.
She said the rapid emergence of AI, combined with widespread digital illiteracy, has worsened the problem. AI-generated content spreads easily on social media and is often believed by large audiences.
Addisalem said she has encountered cases in which even survivors were unaware they had been targeted by AI-generated content because of its high level of sophistication.
Generative AI, Veo 3 and Misogynistic Tropes Against Ethiopian Women
The emergence of advanced generative tools such as Veo 3 has significantly altered the digital playing field for women in Ethiopia. Such technology can generate ultra-realistic videos that exploit cultural sensitivities around modesty, tradition and social norms to spread misinformation and turn women into scapegoats. The power of public embarrassment is being used to undermine and silence women in online public spaces.
The four videos below were the most prominent examples monitored over a five-month period between June and December 2025. The monitoring focused on Veo 3-generated videos targeting women, and selected the four with the highest engagement.
Most of the videos were generated in Amharic, one of Ethiopia’s most widely spoken languages and the federal working language.
Video #1
The video shows a hyper-realistic nightclub where an elderly man is surrounded by several young women. Using Veo 3’s synchronized audio, the man delivers a provocative monologue suggesting that wealth grants sexual access to young women regardless of age, even requesting “Viagra” from the audience. The content is a textbook case of gender trolling. It is not just a joke; it is designed to validate the “materialistic woman” trope. By creating realistic-looking evidence of this behavior, the video incites gendered hate and justifies the harassment of young women who participate in digital spaces.
In the Ethiopian context, the video targets ideas of modesty and traditional values. By fabricating a scene of blatant sexual transactionalism, it triggers shaming against young women, feeds into manosphere rhetoric that labels women as “gold diggers,” and creates a chilling effect in which young women may avoid public spaces or social media to escape being associated with such manufactured trends.
A renowned journalist and feminist activist, who requested anonymity because of her current circumstances, said the video portrays women as being attracted to any man with money, reducing their agency and reinforcing harmful stereotypes.
She noted that the older man is depicted as powerful and dominant solely because of his wealth, while the younger women are portrayed as dependent on him, further entrenching unequal power dynamics. The depiction of two women competing for the attention of a misogynistic man, she added, amplifies toxic gender norms — patterns that AI systems may inadvertently perpetuate by promoting and normalizing such content.
She warned that when AI-generated videos are used by malicious actors, the potential scale of harm could be immeasurable.
Ruth, whose name has been changed for safety reasons, says she no longer feels secure on social media.
“Even though I try to resist intimidation, slurs and other forms of online harassment, I don’t feel immune to Veo 3 videos that could spread unfounded claims about me,” she says.
Her concern extends beyond her own safety. Ruth worries about how such content could affect her family.
“If my family were to see misrepresentations or completely fabricated videos about me, it would devastate them,” she says. “That’s why I stay cautious and keep a low profile online. I avoid drawing attention to myself or engaging in arguments, even when I have ideas I want to share.”
Ruth believes the threat reflects a broader pattern of silencing women online. According to her, the fear of harassment and manipulated videos discourages women and girls from expressing their views as openly as men on social media.
“The environment makes social media increasingly hostile to women,” she says, adding that the growing spread of Veo 3-generated content has heightened fears of potential attacks and discouraged women’s participation online.
Video #2
The video depicts a traditional wedding ceremony, complete with realistic lighting, guests and attire, between a centenarian man and a 25-year-old woman. The central hook is the dialogue. When asked their ages, the man identifies his age as “100 million,” referring to his net worth, and the crowd applauds and celebrates.
By showing the crowd clapping and celebrating, the AI generates false social proof. It makes it appear that society, and the woman in the video, fully endorse the idea that a woman’s presence and love are purely for sale. The young woman is stripped of agency and presented as a prize for the man’s wealth. When AI models use facial features common to Ethiopian women, they create a universalized victim, allowing any young woman to be projected into this submissive, materialistic role.
In Ethiopia, marriage and family values are highly regarded. This video weaponizes those values by invoking public outrage. The “100 million” age joke mocks the economic struggles of young men and redirects their frustration toward women.
It portrays the exploitation of age gaps as a celebrated financial transaction, which can desensitize the public to actual cases of forced or economically coerced marriages.
It also contributes to a toxic online environment in which Ethiopian women are pre-emptively judged as materialistic, making it harder for them to use digital platforms for professional or personal growth without facing “gold digger” tropes.
Video #3
The video begins with a realistic street interview in which a woman is depicted in a highly vulnerable and non-consensual state, rendered half-naked. The male interviewer makes a direct sexual proposition. The woman’s AI-generated response reinforces the materialistic trope: she conditions her consent on the man owning a car and a home. The video then uses a seamless AI transition to move the setting into a public taxi, where the man is replaced by a chimpanzee. When the animal claims to own a car, the woman approves of the sexual encounter.
Using AI to swap a human male for a chimpanzee is a tactical move intended to suggest that women have lower standards than animals when money is involved. This is a severe form of gendered hate speech. Even if the woman is synthetic, the AI is trained on real human likenesses. Generating a woman in her underwear in a public street setting without consent is a form of image-based sexual abuse.
In a conservative society such as Ethiopia, depicting a woman half-naked in a public space, even a synthetic street, is an act of extreme social violence intended to cause social death. By suggesting that a woman would choose an animal over a man if the animal had money, the video attempts to justify physical or verbal violence against women by portraying them as subhuman, instinctual and materialistic.
Video #4
The video features a synthetic male figure holding a phone, mimicking a popular reaction style common on TikTok and Telegram. He points to a real, viral photograph of a young Ethiopian woman known for her distinct spiritual and life views. The generated audio explicitly insults her, stating that a person with big eyes is stupid or brainless.
By pulling a real woman’s photo into an AI-generated environment without her consent, the perpetrator is committing a form of non-consensual use of likeness. The video uses AI to revive physiognomy, the discredited practice of judging a person’s character by physical features, in this case her eyes. AI makes this pseudo-science look like a modern, viral truth.
This is a tactic used to punish women who go viral for their ideas or beliefs. The goal is to shift the conversation from her ideology to her appearance, effectively stripping her of intellectual authority.
This video represents a transition from generalized misogyny to individualized attack. It shows how AI is being used not only to create fake stories, but also to actively dismantle the lives of real Ethiopian women who dare to hold and express an ideology.
Gaps in Justice and Accountability
Ethiopia’s justice system is widely viewed as insufficiently responsive to technology-facilitated gender-based violence (TFGBV). AI-generated videos targeting officials are often swiftly contained and removed. By contrast, similar attacks against women and other vulnerable groups are frequently met with hesitation and inaction.
There is also a significant gap in technological understanding within law enforcement institutions. While police officers receive periodic training, Addisalem said it remains insufficient. AI-related training is resource-intensive and currently inadequate. She added that the absence of a clear legal framework addressing TFGBV further complicates accountability and effective response.
Policy Claims and Contestation
Recently, Prime Minister Abiy Ahmed announced that Ethiopia plans to launch an AI university, which the government says would make the country the second in the world to establish a specialized institution focused exclusively on artificial intelligence, including machine learning and robotics. The prime minister also said the country has a clear framework governing AI.
However, various commentators dispute this claim, arguing that the absence of a legal framework makes it impossible to effectively mobilize resources to address AI-driven misinformation, TFGBV, hate speech and related harms.
A group of users reported all the videos discussed above to TikTok. According to those involved in the reports, the platform responded that the videos did not violate its guidelines and declined to take further action.
Veo 3, owned by Google, appears to be the major tool used to produce such content. The cases raise questions about whether the tool has safeguards robust enough to protect vulnerable groups, particularly in countries such as Ethiopia, where TFGBV is rampant and platform moderation often struggles with local language and context.
Conclusion
Technology-facilitated gender-based violence is increasing sharply, and the emergence of AI has made it more complex and harder to control. The harm does not remain confined to online spaces; it places women’s lives and safety at risk and often escalates into real-world consequences.
While AI tools, particularly platforms such as Google’s Veo 3, offer significant technological potential, they are also highly vulnerable to misuse. Online violence is now increasingly facilitated by AI, further exposing already vulnerable segments of society. AI-generated abuse can cut across class, race, religion and gender. However, the response to online violence against women remains minimal, demanding stronger and more deliberate action from policymakers, justice institutions and technology platforms.
The uneven protection against technology-facilitated gender-based violence in Ethiopia is evident. This imbalance has allowed the problem to persist and expand, increasing its scale and deepening its impact on survivors.
The lack of language versatility on major platforms, particularly in the Ethiopian context, means a significant amount of harmful content targeting women continues to circulate widely on social media, regardless of intent or impact. Because platforms such as TikTok, Facebook and others widely used in the country have limited capacity to process local languages, they face considerable challenges in accurately evaluating content, including context, intent and potential harm.
