The ownership mess of AI-generated art

AI art sounds like a fascinating development, but it puts real artists at risk

pixelated sketchy illustration of a face
ILLUSTRATION: Youngin Cho / The Peak

By: Daniel Salcedo Rubio, Features Editor

Artificial intelligence (AI) is slowly being adapted into our everyday lives. We see it used for relatively simple things, like the Netflix recommendation engine, digital assistants like Siri and Cortana, and image generators. These generators use programming called “deep learning” — the gist is that the software mimics a biological brain. Its developer gives the AI a set of images to learn patterns from, so if you give it a set of landscape photos, it will eventually recognize what sky and mountains are, and use this knowledge to create entirely new images. While this might sound groundbreaking, the current state of image generator AIs is controversial as it puts existing human artists at risk.

I like to consider myself an artist on the side, and I’m protective of the pieces I’ve created. I know this isn’t an isolated concern: look at what happened to artist Tuesday Bassen. Fashion retailer Zara allegedly stole her artwork for their merchandise and profited off her designs. Artwork theft and copyright infringement will likely only get worse with increasing use of AI. Artists have legal ownership over the work they create. Image generators like DALL-E-2 can’t violate this right, no matter how innovative they may be.

Since AIs need a data set to learn and recognize patterns, one major worry artists have is that their artwork is being used for AI learning, raising concerns of copyright infringement. Yes, artists use similar imitation strategies to learn and polish their skills. I myself have a Pinterest account full of reference material from other artists. However, there’s a huge difference between someone referencing other artists’ work to build their own skills, and an AI mixing-and-matching copyrighted material to create a “new” piece of art. Developers must ensure the data set they’re using for deep learning is free from copyright. 

How can we define the ownership of an AI-created piece? Is the person who gives the program its prompts the owner of this “new” creation? Should ownership go to the developer who built the AI? Or, should there be a type of communal ownership by the artists whose work was used for AI learning? There’s technically no right answer to any of these questions; each person involved is a key player in the creation of AI art. And it’s precisely because there’s no right answer that AI image generator users and developers should tread carefully in this new arena. 

Defining ownership is a required step before existing artwork can be used for financial gain. As it stands right now, image generators pose a significant threat to the livelihood of photographers, artists, and anyone involved in image generation. Businesses can use these open image generators to create free content for their own financial gain. In a world where you can create whatever image you can possibly imagine, what’s the financial justification to hire an artist? Not only that, but this also hurts the artists whose work was used to develop the AI. Their personal style which likely took years to create is now in the hands of corporate entities.

AI image generators are a fascinating new development. They’re an entire new tool for artists to work with, and will eventually open new opportunities for the artistic community. But, as is the case with many technological developments, image generators are moving at a faster pace than what society can regulate. We need a distinct  legal definition of ownership for AI generated artwork, and full transparency of the data sets used to develop the AI as soon as possible.