Midjourney might ban Biden and Trump images this election season

With the rise of AI tools that can quickly create modified images and videos, making fake images to spread political misinformation leading to the upcoming US presidential election has become easier than ever. Midjourney’s solution to that might be to ban political images altogether, according to Bloomberg. David Holz, Midjourney’s CEO, reportedly told users during a chat session on Discord that the company is close to banning images such as those of Biden and Trump over the next 12 months.

“I know it’s fun to make Trump pictures — I make Trump pictures,” he told users who attended the session. “Trump is aesthetically really interesting. However, probably better to just not — better to pull out a little bit during this election. We’ll see.” As Bloomberg notes, people had previously used the company’s AI to generate deepfakes of Trump getting arrested. The company ended free trials for its AI image generator after those images — along with those infamous deepfakes of the pope wearing a Balenciaga-inspired coat — went viral.

At the moment, the company already has rules in place prohibiting the creation of “misleading public figures” and “events portrayals” with the “potential to mislead.” Bloomberg was still able to create modified images of Trump covered in spaghetti using the older version of Midjourney’s system, though, whereas the newer version refused to generate modified images of the former President. Of course, even if Midjourney does ban images of high-profile politicians, it will only be protecting its platform from drawing the ire of critics and becoming the center of attention this election season. It will not prevent the use of AI tools in political disinformation campaigns or the spread fake information meant to manipulate the elections as a whole.

Other tech companies have also taken steps to help prevent political disinformation, or at least to help make it easier to identify. ChatGPT will soon start tagging images created using DALL-E 3, while Meta is working to develop technology that can detect and signify whether an image, video or audio clip has been generated using AI.