Midjourney, the company behind the popular text-to-image AI image generator, has suspended access to free trials following the rise in viral deepfakes generated by its technology.
Midjourney founder David Holz wrote on the company’s Discord server earlier this week:
A “deep fake” is media that looks realistic, created by AI altering an image, video, or audio to show a person doing or saying things that a real person has never done. . Deep fakes can be used for fun and entertainment, but the technology can also spread misinformation, influence opinions, and damage reputations.
Earlier this week, a photo of Pope Francis wearing an ultra-stylish white puffer jacket and diamond-encrusted necklace hit the web. When mainstream news sites reported last week that former President Donald Trump was about to be arrested in New York, photos of the former president being chased by police went viral. The images seemed more believable than similar types of fakes professionally created using Photoshop.
Many generative AI sites, including Midjourney and Playground AI, already have restrictions on prompts containing sexual content. However, all of these sites produce photos that are not real, so the question of what use is acceptable is still debated as far as deep fakes are concerned.
Mid-journey subscription plans range from $10 to $60 per month, depending on the processing power and features you need (such as the ability to run concurrent jobs). The platform also offers an annual subscription plan.
It is unclear how closing the free tier will resolve this issue. Midjourney’s terms of service already prohibit offensive images of celebrities and public figures.
Decryption has reached out to Stable Diffusion creators Midjourney and Stability AI for comment, but has yet to hear back.