Midjourney didn’t flinch at fake Trump images, but platform abuse motivated free trial halt

Deepfake photos are fast approaching the point where many can’t tell the difference between them and real images. It’s a lesson AI image generator Midjourney learned the hard way when some used its platform to create several recent images, including several fakes of former US President Donald Trump being arrested. After the images went viral, many thought they were real, sparking conversations and controversy based on completely fabricated information.

According to The Washington Post and reported on by The Verge, these deepfakes became so widespread, they prompted Midjourney to halt free trials on March 28. However, when TechRadar reached out to Midjourney CEO and founder David Holz for clarification, he insisted that there was a different reason for the halt: “Users signing up for multiple free trial accounts was bringing down the service for paid users so that’s why we temporarily halted free trials.”

This correlates with a previous statement Holz made to The Verge concerning the free trials, that it was due to “extraordinary demand and trial abuse.” Previously, Midjourney’s free trial allowed for 25 free images a month before users had to pay at least a $10 per month subscription fee. 

When asked about the creation of these controversial deepfakes created through Midjourney, Holz told TechRadar “I think we’re still trying to figure out what the right moderation policies are. We are taking feedback from experts and the community and trying to be really thoughtful. We already have some new systems coming which should ship soon.”

Now, though, it looks like Midjourney will be keeping free trials disabled, at least for the immediate future. “We tried turning trials back on again with new safeties for abuse but they didn’t seem to be sufficient so we are turning it back off again to maintain the service for everyone else,” Holz told us.

He added that free trials will be coming back eventually, but the company is still considering how to go about it. “We’re still trying to figure out how to bring free trials back, we tried to require an active email but that wasn’t enough so we’re back to the drawing board.”

What is Midjourney?

Midjourney is just one of several AI image generators, with others including Dall-E, Jasper Art, Starry AI, Stable Diffusion, and plenty of others. These sites exist to create new and entirely fabricated images that are trained on existing online images and data. The images can be photorealistic, fanciful images of animals, surreal images, pixel art, and more. What originally set Midjourney apart from other generators was the more painting-like quality of its images compared to others. At least until version 5 was released on March 15.

Initially, Midjourney-created images were easily spotted thanks to common signs like wonky hands (too many fingers). But once version 5 of its software launched, images became vastly superior in quality, which meant that people could create images with more realistic hands, as well as better lighting and fabric. Essentially you can make tons of deepfakes of celebrities and other public figures thanks to the combination of better technology and the sheer number of real-life images available for training Midjourney.

Deepfakes, or photos created using AI technology for the purpose of spreading misinformation with fake images, have become a growing problem. Beyond the Donald Trump arrest deepfakes, others created with Midjourney featured Pope Francis wearing a puffy, white down coat, French President Emmanuel Macron walking through the ongoing protests in Paris, and Twitter CEO Elon Musk holding hands with U.S. representative Alexandria Ocasio-Cortez.

This isn’t the first time that AI-generated creations have been at the center of controversy, as tech leaders and AI experts recently released an open letter calling for the halt of training AI chatbots like ChatGPT. 

The future of AI-generated image sites like Midjourney is uncertain. While the technology holds great potential to aid humans, it can and has already been used in more destructive ways. Only time will tell whether safeguards, from lawmakers to the companies themselves, can manage safety and public good, while still allowing AI to flourish and grow.