Midjourney recently took action by banning all employees from Stability AI from its service indefinitely.

This decision was made after detecting suspicious activity believed to be an attempt by a Stability employee to scrape prompt and image pairs in bulk. During the journey, advocate Nick St. Pierre shared news of the announcement, which was made through Midjourney’s official Discord channel.

Instructions are provided in written form (for example, “a cat in a car holding a can of beer”) for generative AI models like Midjourney and Stability AI’s Stable Diffusion 3 (SD3) to create images. Utilizing prompt and image pairs may enhance the training or fine-tuning of another AI image generator model.

Activity from bots occurred around midnight on March 2, resulting in a 24-hour outage for the commercial image generator service. Midjourney connected multiple paid accounts with a Stability AI data team member who was attempting to collect prompt and image pairs. Midjourney decided to indefinitely prohibit all Stability AI employees from using the service. A new policy has been introduced: “aggressive automation or taking down the service will lead to the banning of all employees from the responsible company.”

Siobhan Ball from The Mary Sue pointed out the irony of Midjourney being upset about their material being scraped, considering they used training data from the Internet without permission. “Generative AI companies are not pleased when images are taken without permission.” Play the tiniest violin in the world.

Midjourney users pay a monthly subscription fee to access an AI image generator that transforms written prompts into vibrant computer-synthesized images. The bot that creates them was trained on millions of artistic works made by people—it’s a practice that some say is disrespectful to artists. In a recent viral tweet, artist Jingna Zhang expressed the profound impact of seeing her name used over 20,000 times in MidJourney. My life’s work and identity boiled down to mere content for a commercial image generator.

Stability Reacts

Following the announcement of the ban, Emad Mostaque, CEO of Stability AI, mentioned that he was investigating the situation and emphasized that any actions taken were not deliberate. He mentioned that it would be wonderful if Midjourney contacted him directly. David Holz, the CEO of Midjourney, mentioned that he has sent some information to assist with the internal investigation in response to X.

Mostaque mentioned in a text message exchange with Reporter that no images were scraped, but a team member had run a bot to collect prompts for a personal project. We’re not exactly sure how that could have caused a gallery site outage, but we apologize if it did. Midjourney is fantastic.

In addition, Mostaque mentioned that his company does not require Midjourney’s data. “We’ve been using synthetic and other data because SD3 outperforms all other models,” he shared on X. When speaking with Reporter, Mostaque also aimed to highlight the differences in his company’s data collection methods compared to those of his competitor. “We only collect data from websites that have a proper robots.txt file and allow it,” Mostaque explains. “I also completed the full opt-out for [Stable Diffusion 3] and Stable Cascade, building on the work done by Spawning.”

Mostaque downplayed the rivalry when discussing Stability’s current relationship with Midjourney. “There isn’t much overlap, but we get along well,” he informed Media, highlighting a significant connection in their histories. “I provided support to Midjourney by offering a cash grant to help them launch, covering the cost of Nvidia A100s for the beta.”