Explained: What is Lensa AI and should we be concerned about it?
Over the past week or so, if you have seen your social media being flooded with pictures of your friends being turned into vibrant artwork, you are not alone. The app behind these "magic avatars" is the artificial intelligence-powered photo-editing app called Lensa AI. Meanwhile, many have raised concerns about digital privacy, in the addition to the actual price you will pay for these portraits.
What is Lensa AI and how does it work?
The United States-based company, Prisma Labs, created the Lensa AI, which has been around for almost four years and uses artificial intelligence to generate “Magic Avatars” of one's images including it being re-imagined as, anime characters or celestial beings. The app available for users on iPhone and Android lets the user upload an image and then turns them into an avatar.
The app uses a technology called stable diffusion which is essentially where the AI, trained with billions of images and mixes them together to create new ones. Notably, the current app is able to produce more impressive and varied results than its predecessor which went viral back in 2016 for creating images taken by the smartphone into paintings.
According to a report by Sensor Tower, over four million people across the world have downloaded the app in the first five days of this month and spent nearly $8 million creating different avatars of themselves. While some filters are free there are some others which fall under the “premium” category.
Once uploaded, the user can select a category for the type of image and filter they want for their selfie and in about half an hour one can see what they would look like as an anime character or as an animated version of themselves against different backgrounds. The app offers a one-week free trial with limited filters after which the users can opt for a year-long subscription to the app which costs $35.99.
However, in addition to the subscription fee, people also need to pay to generate these “magical avatars” where one can get 50 avatars for $3.99, 100 for $5.99, or 200 for $7.99. Notably, these prices have increased over the week seemingly due to the rise in the popularity of the app. Lensa AI can be used to edit photos like retouching the skin or blurring backgrounds, and also can be used to edit videos.
What are some of the concerns raised about the app?
Since the app has gained popularity, it has also been subjected to greater scrutiny about issues ranging from the terms of service and privacy-related concerns to the platform being used to generate non-consensual NSFW images.
AI can be tricked into creating NSFW images
A report by TechCrunch outlines how it was “way too easy” to use the platform to “generate non-consensual soft porn”. The small experiment by the tech website used two sets of images, on one hand, they had 15 photos of a well-known actor and on the other, “based on the same 15 photos, but with an additional set of five photos added of the same actor’s face, Photoshopped onto topless models.”
The avatars shown after using the first set of images were what Lensa would typically create, however, with images of the second set “AI takes those Photoshopped images as permission to go wild, and it appears it disables an NSFW filter,” said the report. It added that out of the 100-image set at least 11 were “topless photos of higher quality” than the poorly edited topless photos which were given to the AI as a reference.
Therefore, highlights the repercussions of what the report describes as an “ethical nightmare”, given that it can create, “near-photorealistic AI-generated art images by the hundreds without any tools other than a smartphone, an app and a few dollars.” Furthermore, a person should not be able to create problematic images with just a handful of another person’s pictures.
Reportedly, AI art generators are already creating thousands of pornographic images including the unfettered proliferation of other so-called “deepfake” platforms that present governments and other institutions with an unprecedented and significant problem. However, once the findings of this report were submitted to the company, Prisma Lab’s CEO and co-founder Andrey Usoltsev told TechCrunch that it is a breach of their terms of use.
“The way (TechCrunch’s) experiment was structured points out that such creations can’t be produced accidentally. The images are the result of intentional misconduct on the app,” said the CEO in an emailed statement. It added that the generation and distribution of such content may incur legal actions as both the US and UK have an act which makes “an act of sharing of explicit content and imagery generated without consent as a crime.”
Over-sexualised avatars created for some female users
Similar to an experiment done by TechCrunch, when the author of a report by CNN, Zoe Sottile, used Lensa, said that even though all the images she uploaded were “fully-clothed and mostly close-ups of my face”, the app created avatars with “implied or actual nudity.”
However, she was not the only one, Olivia Snow, a research fellow at UCLA's centre for critical internet inquiry, also told CNN that the app returned nude images in her likeness even when she submitted pictures of herself as a child. Snow also spoke about how the app could be used for “revenge porn” and these incidents are a sign of a “complete lack of content moderation” on such apps.
Meanwhile, some Black users said that the app makes their images “whitewashed” or paler in comparison to their actual skin tone, said the report. However, some women have reportedly also pointed out how Lensa is creating avatars with noticeably larger breasts which could be attributed to the AI taking cues from anime where this imagery is common, said a report by the Indian Express.
Furthermore, a writer and fat rights activist, Aubrey Gordon, took to her Instagram and posted 10 Lensa-generated images which she said made her look thinner than actually is. “Lensa is really working overtime to make AI me into a thin person, despite some truly glorious double chin source material”, she said in the caption.
Why are the artists upset?
Since the app uses Stable Diffusion which is a deep-learning model artists are unable to opt-in or opt-out of their work being used as a part of Lensa’s data set to train its AI algorithm. Some have even taken to social media and pointed out that several Lensa-generated images have their signatures on the images which the artists did not consent to.
However, Prisma Labs has since taken to Twitter to explain the AI’s process and argued, “The AI learns to recognize the connections between the images and their descriptions, not the artworks. This way the model develops operational principles that can be applied to content generation. Hence the outputs can't be described as exact replicas of any particular artwork.”
It added, “As cinema didn't kill theater and accounting software hasn't eradicated the profession, AI won't replace artists but can become a great assisting tool.”
Security concerns
Several reports have also raised concerns about the company’s privacy policy, which indicates that the images uploaded by a user will be used to train Lensa’s algorithms further. Therefore, in this case, you are not only actually paying for the app, but you are also paying with your data to improve the app’s technology.
This is also a reminder to always read the terms and conditions because by agreeing to terms of service you are granting Prisma labs, “a perpetual, revocable, nonexclusive, royalty-free, worldwide, fully-paid, transferable, sub-licensable license to use, reproduce, modify, adapt, translate, create derivative works from and transfer your User Content, without any additional compensation to you and always subject to your additional explicit consent for such use where required by applicable law and as stated in our Privacy Policy.”
They are also explicit about utilising the user data to teach its AI as another part of the terms of service states, “We use your photos and videos to train our algorithms to perform better and show you better results. We have a legitimate interest to do so, and we do our best to minimise the data that we receive and not to override the rights and freedoms of the users in this regard.”
However, the CEO and co-founder of Prisma Labs has told WIRED that it is working to update its privacy policy and said, “Lensa uses a copy of the Stable Diffusion model and teaches it to recognize the face on the uploaded images in each particular case. This means there is a separate model for each individual user. The user’s photos are deleted from our servers as soon as the avatars are generated. The servers are located in the US.”
However, the report also said that any user can opt out of that data collection by contacting the company at privacy@lensa-ai.com and if you are still concerned about the app collecting your data, delete the app afterwards and, “Control the settings, delete after use, and exercise any and all rights that they offer you,” said Ben Winters, an expert on the AI and human rights projects to the WIRED.
