People who created deep fakes of Taylor Swift thought it was a game, new unravellings in the saga
2 min. read
Published on
Share this article
Improve this guide
Read our disclosure page to find out how can you help MSPoweruser sustain the editorial team Read more
Key notes
- AI-generated Taylor Swift images on social media traced to the 4chan challenge.
- 4chan users gamify AI image creation, testing tool limits and sparking viral spread.
- Graphika identifies the 4chan thread encouraging safeguards evasion in image generator tools.
The research firm Graphika traced explicit Taylor Swift images back to the 4chan message board, notorious for hate speech, conspiracy theories, and offensive content created using AI. For several days, X blocked searches for Taylor Swift.
But first, what’s 4chan?
4chan is an English-language imageboard website that was launched in 2003. It is famous for its various boards that cater to different topics, including anime, manga, video games, and politics. Users on 4chan post anonymously, using a unique ID number instead of a username.
This anonymity has contributed to the site’s reputation as a breeding ground for offensive and sometimes harmful content. However, it has also been a launchpad for many internet memes and online movements.
According to Graphika, as reported by The NY Times, users on 4chan created these synthetic images as part of a game or were challenged to test if they could produce lewd and sometimes violent images of well-known female figures, including Taylor Swift. The images spread across various platforms, garnering millions of views, leading to Swift’s fans’ defense and calling for stronger protections against A.I.-created content.
The 4chan thread encouraged users to bypass safeguards from image generator tools like OpenAI’s DALL-E, Microsoft Designer, and Bing Image Creator. The activity was characterized as a gamified form of malicious content creation, where users sought recognition and approval for their ability to evade restrictions.
Graphika emphasized that Taylor Swift wasn’t the only victim in this 4chan community, as many other actresses, singers, and politicians were targeted. OpenAI clarified that the explicit images of Swift were not generated using its tools and mentioned safety measures, such as content filtering during training.
Microsoft stated it was investigating the images and had strengthened safety systems to prevent misuse of its services. The text also mentions the longstanding issue of fake pornography generated with software, affecting various individuals since 2017. Microsoft earlier revealed that they have fixed the loophole that created the deepfakes.
More here.
