Midjourney accused “botnet-like activity” by two separate accounts associated with the founder of Stable Diffusion to an approximately 24-hour service interruption. Staff members of competitor generative AI startup Stability AI are allegedly responsible for a systems failure earlier this month to collect Midjourney’s data, as reported by Midjourney, which has restricted Stability AI employees from accessing its service since then.
On March 2nd, Midjourney announced on their Discord support that created photographs were not showing up in the customer gallery due to a long-running server breakdown. In a summation of a March 6th business overview call, Midjourney stated this service disruption was caused by botnet-like activity from premium accounts directly connected to Stability AI workers.
The system was shut offline on Saturday night, as reported by Midjourney member Nick St. Pierre on social media platform X, who heard the call. Midjourney stated that a Stability AI professional attempted to take all the prompting and imagery combinations. Following St. Pierre, a Stability AI data department member had several paid accounts connected to them by Midjourney.
Midjourney describes the company’s overview call on March 6th as office hours. In reaction to the disruption, the company claims it prohibits all Stability AI workers from utilizing its service in the future. Additionally, Midjourney is adopting an updated guideline that will prohibit workers from working for any business that engages in excessive automation or results in service failures.
When St. Pierre brought up the allegations, Emad Mostaque, the chief executive officer of Stability AI, responded on X, stating the company was looking into the matter and confirming that Stability wasn’t providing the command for the acts in inquiry. How the two accounts would do so is still being determined. Since we began utilizing synthetic and other data and SD3 exceeds the rest of the models, this team needs to be scratching, Mostaque stated, pointing towards the Stable Diffusion 3 artificial intelligence model presently in testing. He remarked that If a Stability employee was the cause of the disruption, it was mistaken and never a distributed denial of service (DDoS) attack.
With the same thread, David Holz, the creator of Midjourney, replied to Mostaque, saying he had sent him a few details to assist with an internal inquiry.
Other than that, the matter remains evolving, and from that March 6th chat, no new information has been given. As of this writing, the Verge has yet to receive a response from Midjourney or Stability AI on their request for a response.
It seems strange that an interruption in service this long could have been caused by scrape actions from only two accounts. Internet creatives have also pointed out the ridiculousness of this scenario, criticizing both businesses and generative artificial intelligence in general severely for using large amounts of unconsented internet data to educate their models. Both Midjourney and Stable Diffusion have been the subject of multiple copyright disputes. In December, the latter was charged with building a creative database to assist with training objectives.
