A recent study conducted by Graphika, a firm that analyzes social networks, has revealed that the fake images of Taylor Swift that went viral on social media in late January were likely a result of a chatroom challenge to bypass filters designed to prevent the creation of AI-generated pornography. The images of the pop star can be traced back to a forum on 4chan, an online image bulletin board notorious for sharing controversial content, including conspiracy theories and hate speech.

According to Graphika's report, users on 4chan created these images of Taylor Swift as part of a game to see if they could produce explicit and sometimes violent visuals of famous women, ranging from singers to politicians. The firm discovered a message thread on 4chan that encouraged users to find ways to bypass the guardrails established by AI-powered image generator tools such as OpenAI's DALL-E, Microsoft Designer, and Bing Image Creator. However, representatives from 4chan, OpenAI, and Microsoft have not yet responded to requests for comment.

The phony images quickly spread to other platforms, garnering millions of views and leading to X (formerly known as Twitter) temporarily blocking searches for Taylor Swift. In response, her dedicated fanbase launched a counteroffensive on the platform, flooding it with positive images of the pop star and using the hashtag #ProtectTaylorSwift. The Screen Actors Guild expressed their concern over these fake images, describing them as upsetting and harmful. They called for the illegalization of the development and dissemination of non-consensual explicit images, particularly those of a lewd nature.

The creation of fake pornographic content using software has been a longstanding issue, with limited regulations leaving victims with little legal recourse to have the images removed. However, the emergence of generative AI tools has further fueled the creation and dissemination of pornographic “deepfake” images, including those featuring celebrities. Moreover, artificial intelligence is also being utilized to target celebrities in other ways, as demonstrated by an AI-generated video in January that featured Taylor Swift's likeness endorsing a fake Le Creuset cookware giveaway. Le Creuset issued an apology to anyone who may have been deceived by the video.

The report by Graphika sheds light on the concerning trend of using AI to create and spread deceptive and harmful content, particularly targeting high-profile individuals like Taylor Swift. The incident has sparked a debate about the need for stronger protections against AI-generated images and the regulation of such content. As lawmakers and advocates push for stricter measures, it remains to be seen how the battle against AI-generated disinformation will unfold in the future.