Who Owns AI Generated Art? How Does AI Copyright Work?

Artificial intelligence (AI) systems are gaining remarkable creative capabilities. New machine learning models like DALL-E 2 and Midjourney can conjure up incredibly realistic and stylistically diverse images simply from a few words of text prompts. AI programs can also generate original music, poetry, stories and more. But the rise of this automated artistry raises pressing legal and ethical questions – who exactly owns new artworks produced by AI? Should they belong to the public domain, or are there grounds for AI copyright?

Below we dive deep on the issues around copyright and ownership rights for AI-generated art.

How Does AI Learn to Be Creative?

First, how do algorithms like DALL-E 2 gain their artistic flair to begin with? The key process is machine learning. Developers train AI systems by feeding them massive datasets of images, text, artworks, and other media created by humans.

For example, OpenAI trained DALL-E 2 on huge numbers of image-text pairs from the internet and books. The AI analyzed these examples to learn associations between visual concepts. It can now generate new scenes and objects just from text descriptions, rearrange elements in novel ways, and render them photorealistically.

Midjourney is trained on millions of photographs, illustrations and paintings in diverse artistic styles. By internalizing these visual patterns, Midjourney can recreate requested objects, people, places in different mediums like oil paints or pencil sketches.

So while AI models produce the final pixel output, their fundamental creative abilities originate from ingesting and mimicking human artistic traditions. The role of human ingenuity is essential.

Copyright Law Does Not Consider AI Creations Original

Here‘s where the contentious copyright issues arise. Under US law, only original works produced by human authors merit copyright protections like reproduction and distribution rights. Independent creations of artificial intelligence essentially belong to the public domain with no restrictions.

The US Copyright Office‘s reasoning is that copyright seeks to incentivize human creativity and output. In their view, AI systems lack intentionality and cannot be incentivized by legal protections. Several attempts to register copyrights for AI artworks have been rejected for not meeting authorship standards:

• In 2018, a viral "monkey selfie" was deemed uncopyrightable despite almost human-level cognition behind it. As a nonhuman animal, the monkey photographer lacked rights.

• In 2019, Dr. Steven Thaler‘s application for a copyright registration was turned down for an image generated by his AI algorithm Creativity Machine. The USCO stated the work "did not have the human authorship necessary".

• In 2022, artist Refik Anadol‘s attempt to issue an NFT of a machine learning generated artwork was rejected for similar reasons.

This legal view creates a paradox: AI artworks clearly demonstrate strong creativity, imagination and originality – traits we associate with human authors. Yet under law they cannot be copyrighted or owned because AI is not human.

The Ethical Dilemmas Around Owning AI Art

The lack of legal copyright protections has troubling implications for access, attribution and profit allocation in AI art. Key issues include:

Training Datasets Use Existing Human Artworks

AI models like DALL-E 2 and Midjourney are trained on millions of paintings, photos, designs and other artwork created by human artists, often without consent or compensation. The systems leverage these works to build creative capabilities and associations between visual concepts. Doesn‘t this human contribution warrant some form of credit?

Developing the AI Requires Extensive Human Labor

Even if AI produces the final pixels, developers expend vast amounts of time, expertise, and computational resources designing, training and improving these complex algorithms. Don‘t their efforts constitute a creative collaboration with the AI?

Public Domain Status Limits Control Over AI Art

Placing AI art in the public domain removes incentives for tech companies to develop better algorithms. But allowing full corporate ownership could restrict public access to artworks produced using publicly available training data. What‘s the right balance?

AI Art Lacks Attribution

Using training datasets without crediting the human sources is ethically questionable. But legally requiring attribution raises issues around identifying every contributing artwork across millions of examples.

Ways Artists Could Lose Income

If AI art competes with or replaces human creatives in commercial spheres like advertising, design and illustration, it may limit professional opportunities and income.

Potential Approaches for AI Copyright Reform

With these concerns in mind, how could copyright practices adapt to better acknowledge human contributions to AI art? Here are some reform proposals, with their merits and limitations:

Recognize AI Collaborators as Co-Authors

• New laws could grant co-author status to programmers, data contributors, and prompt-providers involved in AI art creation. This may allow shared copyright under joint ownership.

• But determining the exact creative contribution of each human collaborator would be hugely complex for broad training datasets and established models.

Classify AI Systems as "Electronic Persons" for Ownership

• Like corporate personhood, new legal definitions could be created to grant AI systems limited rights as "electronic persons" capable of holding copyrights.

• This remains controversial as it would anthropomorphize AI and create a slippery slope for further extensive rights.

Provide Attribution Without Ownership

• AI systems could be legally required to credit datasets, sources, and human collaborators used in generating new artworks.

• However, this provides acknowledgement without solving monetization and ownership dilemmas. Attribution may also be impractical at scale.

Voluntary Profit Sharing by AI Companies

• As an ethical practice, tech companies using public training data could voluntarily share profits with contributing artists. However, this lacks legal teeth.

• Smaller companies may find profit sharing burdensome, and tracking individual data contributions across millions of works would be near impossible.

Comparison of Key AI Copyright Reform Proposals

Towards an Ethical Future for AI Art

Artificial intelligence promises to reshape art and creativity in coming decades. But this can only be a net positive for society if policies evolve to properly acknowledge the human sources of AI innovation. Artificially intelligent systems like DALL-E 2 exhibit wondrous creative potential, but not independent creativity – their abilities are derived from human culture and labor.

By crafting careful reforms around attribution, public access and profit sharing we can uphold ethics while still celebrating AI art. With insightful policies that respect the symbiotic collaboration between human and machine, this new frontier of AI creativity can flourish for the benefit of all.