AI-generated nudes of Taylor Swift are appearing all over the internet, and there’s no evidence that they will be stopping anytime soon.
The images are spreading across social media platforms after seemingly being posted to a celebrity nudes site, which placed its watermark in the corner of each image. And as always, the tech companies behind those platforms are struggling to crack down on the abuse.
Videos by VICE
On X—which used to be called Twitter before it was bought by billionaire edgelord Elon Musk—the account which initially posted the AI nudes has been suspended. But the images are still widely available via a quick search of the platform—as are various “nudify” apps that allow users to virtually undress women with generative AI. On Thursday afternoon, the Swift images were still being shared widely by various accounts with blue checks—a meaningless label that used to indicate verified accounts, but is now given to anyone who pays to subscribe to Musk’s platform.
This is not especially surprising given that Musk has virtually eliminated the platform’s moderation staff since buying the company. Other platforms have followed suit, eliminating staff positions in charge of combating hate speech and misinformation. X did not respond to Motherboard’s request for comment beyond its automatic boilerplate email reply.
Reddit, which banned non-consensual nude deepfakes after they initially proliferated there in 2018, has also been taking down posts and moderating users who share the images, a spokesperson told Motherboard. The site uses a combination of automated tools and human review to accomplish this; it appears to have largely worked, although Motherboard found examples of the images being shared on small subreddits. The images were also reportedly circulating on Instagram; Motherboard couldn’t find examples of the viral AI nudes in a search, but did uncover other pornographic, non-nude deepfakes featuring Swift. Meta said in a statement that it bans sexually explicit images on its platforms using automated tools.
“This content violates our policies and we’re removing it from our platforms and taking action against accounts that posted it,” the Meta spokesperson said. “We’re continuing to monitor and if we identify any additional violating content we’ll remove it and take appropriate action.”
This was all easy to see coming. Taylor Swift is probably the single most visible woman of this century, so it makes sense the chart-topping pop star would be the most obvious target for AI-generated nastiness. Last month, a viral TikTok scam imitated Swift’s voice and likeness to advertise a fake offer for free cookware sets. The Swift brand is simply too big and too ubiquitous to not be defiled by mouth-breathing internet goblins who get off by publicly fantasizing about having sex with female celebrities.
This doesn’t bode well for other, less-famous women and femme-presenting people, either. In many ways, this is a nightmare scenario for anyone whose bodies are routinely sexualized and exploited—and especially for teenagers who are most likely to be harmed by AI nudes. Most recently, teenage girls in New Jersey reported that bullies had begun spreading AI-generated nudes of them in their school. And there have been various other incidents where abusers have used “nudify” apps to generate explicit pictures of classmates and online influencers.
The sad but entirely predictable proliferation of non-consensual AI porn is just one obvious consequence of the rise of AI-generated content, enabled by large corporations which profit from AI tools with virtually zero accountability. Indeed, deepfakes originated explicitly to create AI porn of real women, a malignant strain of the technology’s DNA that has yet to be excised.
Companies like OpenAI, Meta, and Microsoft have attempted to mitigate the production of harmful content with filters that block users from generating abusive and illegal stuff. But this has proven to be a cat-and-mouse game, largely because the systems themselves are trained on billions of nonconsensually obtained images scraped from the internet. For example, one of the largest datasets used to train AI systems, LAION, was recently found to contain over 3,000 explicit images of child sex abuse. This was met with a complete lack of surprise from AI researchers, who have been warning about this for years. In addition, AI models are far from scarce, and it’s easy for someone with a bit of know-how to tweak an existing one to produce all kinds of abusive and illegal images.
More recently, lawmakers across the US have proposed bills that would explicitly make it illegal to generate nonconsensual AI nudes. But experts note the bills seem mostly geared toward protecting the intellectual property rights of rich celebrities like Swift, and it remains to be seen whether these laws would do anything to protect the rest of us.
Jordan Pearson contributed reporting to this article.
