Holly Rowe isn't alone — AI-generated videos of violence against women are all over the internet

Editor's note: This story contains a detailed discussion of videos depicting violence against women.

An AI-generated video depicting ESPN sideline reporter Holly Rowe being violently slapped by Texas football coach Steve Sarkisian sparked major controversy in the world of sports last week. It's also offered a glimpse into a largely overlooked corner of the internet where fake clips of women being struck, attacked and even killed are rampant.

The AI video of Rowe and Sarkisian stemmed from a real-life interaction between the two after Texas's stunning come-from-behind victory over Ohio State on Saturday. Rather than answer Rowe's post-game questions, Sarkisian said only a few words before walking away, leaving her looking frustrated. A short clip of the exchange was passed around liberally online afterward.

An unknown social media user then augmented that real footage to show Sarkisian striking Rowe, sending her to the ground. Though the fake clip garnered some attention when it was first posted, it became a huge story when Texas's star quarterback Arch Manning was asked about it during a press conference on Monday.

"I saw an AI edit of Sark's interview. Have y'all seen that one?" Manning said , laughing. "That's a good one. Of him slapping Holly Rowe? It's hilarious."

Manning has since apologized to Rowe. In a Sept. 15 post on social media, she accepted his apology and said she is ready to "move forward with respect and kindness for one another."

Although it may have been some internet users' first encounter with this type of AI video, social media sites are rife with content depicting similar — and, in many cases, much more extreme — acts of violence against women.

In just a few short years, AI-powered image and video creators have become remarkably adept at making hyper-realistic content that even those with a sharp eye can struggle to identify as inauthentic.

Despite guardrails put in place by many of the companies that make these tools, they are often abused. This abuse comes in many forms, but the most common and most widely recognized is pornography, including " deepfakes " where a real person's face is transposed into sexual content and AI-generated child sexual abuse material.

Fake videos of non-sexual violence against women are less common, but still easy to find.

"Image-based abuse, whether sexually explicit or not, is a massive category of online violence against women and girls," Adam Dodge, founder of EndTAB , an organization dedicated to ending technology-enabled abuse, told Yahoo News via email. "While the images may differ, the motivations remain the same. They are weaponized to intimidate, objectify, harass, and exert power and control over the person depicted."

The videos come in several forms. Sometimes, it's a short video of an elderly woman being punched by a large male fighter or being inexplicably launched through a door by her chiropractor. These videos are ostensibly intended to be funny, and the absurdity of what they depict often makes them easily recognizable as AI.

But others are much more realistic and disturbing. Last year, the tech news site 404 Media discovered a YouTube channel filled with photo-realistic videos of women being shot. The channel, which has since been deleted, had reportedly uploaded dozens of videos of women begging for their lives before being shot in the head.

"AI-generated depictions of violence against women serve similar purposes that depictions of violence against women have always served – whether in books and magazines, or on screen in movies and TV shows, and now in AI-generated online content," Alexis Champion, A Technology Safety Senior Specialist at the National Network to End Domestic Violence , told Yahoo News. "These depictions normalize the violence and reinforce cultural norms that justify and excuse the oppression and abuse of women and girls."

There is also a relatively widespread trend of realistic AI videos of women being strangled, with. a quick search on TikTok surfacing multiple accounts dedicated almost entirely to these clips. One 15-second video featuring Supergirl being strangled by an unseen attacker has more than 400,000 views on TikTok.

Women being hurt is a regular feature of the endless stream of fake videos purporting to capture real viral moments that have infested online spaces these days. Often presented as footage from a bodycam, security camera or dashcam, these videos variably show women as they're manhandled by police, attacked by animals (including a tiger) or fighting each other in achingly melodramatic scenarios.

"The fact that the videos aren't real doesn't make the fake footage any less harmful or dangerous," wrote the authors of an online petition with nearly 46,000 supporters calling for AI companies to prevent their platforms from making this type of content. "Real brutality against women and girls is bolstered by these fake videos, reinforcing a culture of violence. … The psychological damage on viewers is unthinkable."

Attacks against women are also just a routine part of the general AI slop that has infested online spaces. One Facebook account with 264,000 followers, ironically named Dignity Preserved, has posted well over 100 glossy, hyper-stylized clips of women being hurt with captions like, "My Billionaire Husband Beat Me At A Gala" and "The Mistress Laughed While the Wife Bleeded on the Floor." The channel's videos regularly receive hundreds of thousands of views.

How AI fuels real-life violence against women

These violent videos are just one small example of the ways that powerful new tools have fueled what the United Nations calls "AI-facilitated violence against women."

One global study from 2021 found that 85% worldwide had either experienced or witnessed online violence against women before the AI boom. Now, AI is "enabling that violence to spread further, faster, and in more complex ways," according to the U.N.

The misogynistic culture that AI-generated pornography and violent videos help bolster ultimately leads to more abuse in the real world, experts say.

"What begins online doesn't stay online. Digital abuse spills into real life," Sima Bahous, the executive director of U.N Women, said.

Beyond feeding harmful attitudes toward women, AI has also been harnessed to facilitate abuse in the real world. There are examples of men using AI-powered cameras to stalk their exes, maliciously publishing deepfake porn to destroy women's reputations, and relying on chatbots to validate their dangerous obsessions.

"The problem with AI-generated content is the scale and speed at which it can be produced, and the ability of the content creators to post anonymously and avoid prosecution and accountability," Champion wrote.

Lawmakers from both parties have proposed a number of bills to rein in AI-powered online abuse. The first of those bills to become law was the TAKE IT DOWN Act , which was signed by President Trump in April 2025. The act makes it a federal offense to distribute non-consensual intimate imagery and requires websites and social media platforms to remove any such material within 48 hours if the person portrayed in the content asks them to.

A separate proposal called the No Fakes Act , which was advanced by the Senate Judiciary Committee in June, would broaden restrictions on deepfakes so they apply whenever someone's likeness is used without permission, rather than only when it's used for sexual content.

Currently, there don't appear to be any bills before Congress that specifically target violent AI-generated images or videos depicting fictional characters or otherwise unidentifiable people.

One reason AI-generated violence hasn't inspired the same level of pushback as other sexually explicit content, Dodge argued, is that its harms may not be as obvious to the casual observer — particularly when it may not necessarily violate the rules of platforms that host it.

"Non-explicit image-based abuse is also uniquely challenging because it may not violate traditional content guidelines, policies and procedures, making it hard to take down," he wrote. "I think it's challenging for the average person to understand why sharing a non-explicit image would be harmful."

Less than a third of people working in AI are women, according to a 2023 report from the World Economic Forum . Champion argues that closing that gender gap could make platforms more likely to root out violent content from the inside.

"Women are significantly underrepresented in AI development and leadership," she wrote. "Women's voices and leadership must be brought into these spaces in much greater numbers in order for effective and relevant measures to be taken to curb the spread of violent content."