On Wednesday, Midjourney indefinitely banned all employees from image synthesis rival Stability AI from its service after it detected “botnet-like” activity that it suspects is causing Stability employees to prompt and Trying to scrape a large number of image pairs. Midjourney’s lawyer Nick St-Pierre tweeted about the announcement, which came via Midjourney’s official Discord channel.
Prompts are written instructions (such as “a cat in a car with a can of beer”) that creative AI models such as Midjourney and Stability AI’s Stable Diffusion 3 (SD3) use to synthesize images. Having instant and image pairs can potentially help in training or fine-tuning a rival AI image generator model.
The bot activity, which occurred around midnight on March 2, caused a 24-hour outage for the commercial image generator service. Midgerny linked several paid accounts with an employee of the Stability AI data team trying to “get prompts and image pairings.” Midjourney then decided to ban all Stability AI employees from the service indefinitely. It also hinted at a new policy: “Aggressive automation or termination of service will result in the ban of all employees of the responsible company.”
The middle journey
Siobhan Ball of The Mary Sue found it ironic that a company like Midjourney, which took its AI image synthesis models using training data off the internet without permission, would be sensitive about scraping its content. “It turns out that creative AI companies don’t like it when you steal, apologize, scratch images from them.
Midjourney users pay a monthly subscription fee to gain access to an AI image generator that turns text input into lush computer-synthesized images. The bot that creates them was trained on millions of artworks created by humans – a process it claims disrespects artists. “Words can’t describe how inhumane it is to see my name used 20,000+ times in Mid-Journey,” artist Jingna Zhang wrote in a recent viral tweet. “My life’s work and who I am – a commercial image reduced to meaningless fodder for a slot machine.”
Stability responds.
Shortly after news of the ban emerged, StabilityAI CEO Imad Mossadegh said he was looking into it and claimed that what happened was not intentional. He also said that it would be great if Midjarni contacted him directly. In a response to X, Midjourney CEO David Holz wrote, “Sent you some information to help with your internal investigation.”
In a text message exchange with Ars Technica, Mostaque said, “We checked and there was no scraping, there was a bot running that was being run by a team member that was personal. Was collecting hints for the project. We’re not sure how it’s going to go. It’s going to cause the gallery site to shut down but sorry if it does, Midgerni is great.”
Besides, Mustaq says, his company doesn’t need Midgerni’s data anyway. “We’re using synthetic and other data that makes SD3 better than all other models,” he wrote on X. In a conversation with Ars, Mostaque similarly wanted to compare his company’s data collection techniques to those of its competitors. “We only scrape what has the proper robots.txt and is allowable,” says Mustaq. “And also did a complete opt-out. [Stable Diffusion 3] And spawning did the trick to take advantage of the stable cascade.”
When asked about the stable relationship with Midjarni these days, Mossadeq dismissed the animosity. “No real overlap, although we get along,” he told Ars, emphasizing a significant link in their histories. “I funded the midjourney to get it. [them] Off the ground with a cash grant to cover [Nvidia] A100s for Beta.”