Midjourney, one of the most widely used A.I.-powered image generators, finds itself at the receiving end of criticism on social media. This comes after Midjourney imposed a ban and changed the terms of service (TOS) when concerns about infringement were raised.
A new version of the AI-image generator called Midjourney V6 was released, touted to be the most advanced to date. However, some users have noticed that it can recreate potentially infringing AI images even with the simplest of prompts. One specific user who has a paid subscription to the Midjourney service was even banned for allegedly violating the TOS.
Reid Southen (username @Rahll on X.com, formerly Twitter) is a professional concept artist. He has worked on such projects as Transformers, The Matrix Resurrections, The Woman King, The Hunger Games, and Blue Beetle.
Southen shared his experience of Midjourney banning his account and deleting all the previous images he generated using the service.
I wasn't going to do this, but because @midjourney nuked my account for exposing their copyright infringement and plagiarism, here's an 11 minute video of me scrolling through all the offending generations on Discord. Enjoy. pic.twitter.com/bW4lszGzwh
— Reid Southen (@Rahll) December 23, 2023
This was allegedly Midjourney’s response after Southen pointed out how Midjourney V6 generated images that were identical to scenes from films. Southen provided examples that were comparable to scenes from the MCU, DCEU, and Dune. According to his posts, it did not require specifically including in the prompts which scenes to recreate.
It did not take long for Southen’s tweets to reach other users who had similar concerns about the potentially infringing images generated through Midjourney. Plus, many agreed Midjourney’s reaction of banning Southen without full disclosure as to what violation he committed for reporting the issue of infringement was counterproductive.
Speaking with Reid Southen himself, he states he stands by what he shared on his tweets regarding the experience with Midjourney. He also added:
“I just think it’s worth noting that I’m not the only person posting images of other people’s IPs using Midjourney, but they chose to ban me specifically because I was critical of them. It’s not the first time this has happened.”
In one of his follow-up tweets, Southen highlighted how Midjourney changed their terms of service (TOS) within 48 hours of his reporting the possible infringing images and subsequent ban of his account.
Midjourney's infringement and conduct have gotten much, much shadier.
V6 released on the 21st, so Midjourney added this to their ToS AFTER the backlash, likely sometime on the 22nd or 23rd when people like me started posting incriminating stuff.
This is insane… pic.twitter.com/m4hRGTB20p
— Reid Southen (@Rahll) December 25, 2023
The use of artificial intelligence without regulations has been a serious concern in 2023. The White House and the EU have already taken initial steps in addressing the issue. Aside from the potentially disruptive technology, there is an increasing public awareness of how A.I.-powered companies exploit proprietary and private data without consent or compensation.
This is particularly the case when it comes to how A.I. is being used to exploit children. The matter has even been raised to the U.S. Congress, prompted by concerns of lackluster federal laws in place to address the matter.
Just recently, a Stanford research found an alarming amount of child sexual abuse material (CSAM) and child pornography (CP) in the LAION-5B large dataset. Stable Diffusion and Midjourney, contemporaries in the A.I. text-to-image generation service, have been trained at some level on the LAION-5B datasets.
But the Stanford report states that it is increasingly difficult to determine how much CSAM and CP content are included, due to the lack of disclosure and transparency. Moreover, the researchers at Stanford stated “removing material from the models themselves is the most difficult task”, putting into question if it is at all possible to remove the offensive imagery by piecemeal.
Since the Stanford report was made public, the LAION-5B datasets were taken offline.
big breaking news: LAION just removed its datasets, following a study from Stanford that found thousands of instances of suspected child sexual abuse material https://t.co/mIYh64WuvS
— Samantha Cole (@samleecole) December 20, 2023
How Midjourney Inc. responded to the fair critical and social concerns regarding infringement further casts doubt on the ethics and morality, if not the legality, in using the A.I. image generator.
But what do you think about Midjourney’s response? Is banning users for contacting an A.I.-powered company such as OpenAI or Midjourney for potential copyright and trademark infringements justified? Should paying subscribers avoid speaking about these legitimate concerns for fear of similar repercussions? Do you believe image-generator services should provide full transparency on what data their models are trained on?
