Midjourney ends free trials of its software after users took advantage of the service to generate fake images, better known as deepfakes. The decision comes from the controversies, in which fake images of Donald Trump arrested or Pope Francis walking around in a modern white coat have gone viral. All false, although many users were not able to detect the deception.
Midjourney CEO and founder David Holz announced the change Tuesday on Discord over the “extraordinary lawsuit and abuse of judgment,” according to statements reported by The Washington Post. From now on, users who want to generate images using Artificial Intelligence (AI) will have to pay ten dollars a month for the service.
Holz acknowledges that the new security measures have not been “sufficient” to prevent misuse during trial periods and says they will continue to work to establish content policies to ensure greater control.
The problems with the generation of false images go back a long way. In 2022, the Midjourney CEO justified banning images of Chinese leader Xi Jinping, arguing that his team only wanted to “minimize the drama” and that having access in China was more important than allowing satirical content.
But the problems increase, because Midjourney is capable of creating increasingly realistic images thanks to version 5, which improves lighting, corrects deformities and recognizes countless celebrities and public figures.
“Moderation is hard and we will be shipping improved systems soon. We are receiving a lot of comments and ideas from experts and the community and we are trying to be really thoughtful, â€explains the person in charge of The Verge.
The free use of Midjourney has allowed the generation of fake images such as the alleged arrest of former US President Donald Trump. Its creator was Eliot Higgins, founder of the Bellingcat journalism page, who typed on the platform “Donald Trump falling while being arrested.” He thought “maybe five people” would retweet those fake images. He was wrong, because there were 5.6 million.
Given the success of his creation, he asked Midjourney for images of Trump being detained, in court, upon his arrival at jail, his life as a prisoner and even his prison escape. He acknowledges that he was just “goofing around.”
Pablo Xavier, a 31-year-old construction worker, also created a deepfake. In his case, Pope Francis is seen dressed in an elegant white Balenciaga coat, an image that seems totally real, but is not. This young man also did it for fun.
Midjourney is not the only platform that is having to regulate the use of its software. For example, OpenAI blocks any images of ongoing political events, conspiracy theories, and politicians. It also prohibits hate, sexuality and violence. Stability AI doesn’t allow Stable Diffusion users to copy styles or make images that aren’t safe for work, though it doesn’t specify much else.
Misleading content isn’t the only problem for these AI platforms. There has been a long-standing concern that images are stolen and copyright infringed. For example, Getty Images, one of the world’s largest agencies, has taken legal action against Stability AI on the grounds that it “infringed intellectual property rights, including copyright.”
