Did Stability Just Steal Images From Midjourney? A Major Feud Between the Companies.
A major feud and irony as Midjourney accuses Stability of data theft through so-called data scraping. It's an ironic turn of events, considering how Midjourney built its service on data scraped from across the web. So what drastic measures has Midjourney taken?
In an unexpected move, Midjourney, a leading player in generative AI, has decided to permanently ban all employees of its competitor Stability AI. This action follows the discovery of what's described as "botnet-like" activity, suspected to have been carried out by a Stability AI employee. This activity involved mass collection of prompt-and-image pairs from Midjourney's services, which are critical components for developing and improving AI models that generate images based on text descriptions. This approach, which could be seen as an attempt to quickly improve a competitor's technology at Midjourney's expense, has led to a marked escalation of tensions between the two companies.
Open Data Becomes Vulnerable to This Type of Attack
Prompt-and-image pairs play a central role in the development of AI-based image generation techniques. By using these pairs, companies can train their AI models to understand and improve how they create visual content from text instructions. Midjourney has had a feature that makes both images and prompts publicly accessible, which leaves them vulnerable.
Midjourney's decision to take action against Stability AI highlights the deep importance of this data, as well as the potential risks and consequences of such information falling into the hands of a direct competitor. This incident points to a growing concern within the AI industry over data security and the ethics surrounding the collection of training data.
Midjourney Acted Quickly
In response to the discovered activity, Midjourney has quickly implemented a new policy: companies deemed harmful to their service through automated processes, or that cause the service to shut down, will have their employees banned. This move marks a strict stance against data scraping and sends a clear warning to other players in the tech sector to respect other companies' data and services. It also illustrates the challenges companies face when navigating the balance between innovation and privacy in an increasingly data-driven world.
The Irony of It All
This conflict has sparked discussion about the irony of Midjourney, which itself built its technology to a large extent using data scraped from the internet, now condemning this incident. Comments from industry observers and critics point to the complex ethical questions in AI development, where the lines between innovation, integrity, and ethical behavior can often become blurred. This situation underscores a broader debate within the tech world about responsible data use and respect for creators' rights.
Stability's Defense
In response to the accusations, Stability AI's CEO, Emad Mostaque, has stated that they are investigating the situation and claims their intentions were never to harm or unlawfully exploit Midjourney's data. He emphasizes that the activity in question was intended for a personal project and denies that any image scraping took place.
Copyrights Get Tangled Up – Does It Become an Even Bigger Gray Area?
Companies are stealing data from each other that was itself stolen from copyright holders. There's another example that maybe hasn't gotten as much attention yet but that we believe will soon come to light. Adobe Firefly is said to have been based on the uploads its users have made. Adobe Firefly is trained purely on that data, and they're trying to stay legally clear of it. We tried uploading copyrighted material to Adobe Firefly, along with its metadata. It went through without a problem, which raises quite a few questions about how secure the service really is.