The Impact of Artificial Intelligence on U.S. Copyright Law and Authorship

Artificial Intelligence & Human Authorship Overview

Human Authorship

As a means to encourage and foster the arts the U.S. government sought to create a means of protecting an artist’s ability to control their works by granting ownership over their artistic expressions. Under the Copyright Act of 1976, the U.S. government grants protection to “original works of authorship fixed in any tangible medium of expression.” Important to the issue of artificial intelligence, “authorship” has been historically interpreted as synonymous with human creativity. Courts and the U.S. Copyright Office have long held that only human beings can hold copyrights.

This principle was reaffirmed in several cases, but most illustrated in the famous “monkey selfie” case (Naruto v. Slater), where a selfie that was taken by a macaque was deemed ineligible for copyright because it lacked human authorship. With this limitation extended to the concept of artificial intelligence a similar question arises: If an artificial intelligence system that was created by humans creates a painting or writes a poem, who—if anyone—owns the copyright? Do the human creators of the system own it? Do the artists whose expressions were used to train the system? Do the users of the system who instructed the system to create the work own it? Is the artificial intelligence system itself a novel authorial entity capable or even worthy of ownership?

The U.S. Copyright Office’s Stance on AI-Generated Works

In February 2022, the Copyright Office’s Review Board issued a final decision affirming the refusal to register a work claimed to be generated with no human involvement. A year later, the Office issued a registration for a comic book incorporating AI-generated material.

In early 2023, the Office announced the launch of a broad AI Initiative and issued a statement of policy providing guidance on the registration of works incorporating AI-generated material. In March 2023, the U.S. Copyright Office released guidance clarifying its position: works that are entirely generated by AI, without human creative input, are not eligible for copyright protection.

However, the situation becomes more nuanced when human creators use artificial intelligence as a tool. For instance:

• If a human artist prompts an artificial intelligence system to generate images and then edits or curates the results meaningfully, they may be eligible for copyright protection, but only for the aspects they created.
• Registrations must disclose the role of artificial intelligence in the creation process.

The key factor, as mentioned previously, is human authorship. In other words, a legal analysis of ownership of works which were created using an artificial intelligence tool will turn on how much creative control the human user of the system exercised in creation of the work.

To illustrate this, consider two songs which were created using an artificial intelligence system as a tool by a single user. Song (A) is created simply by the user issuing the prompt “write a song about pirates;” while song (B) is created by the user writing original lyrics about pirates, submitting a chord progression and issuing the prompt “write a 1980’s hair metal song about pirates using only one drum track, one bass track, one guitar track and one vocal track.” While neither example would result in the user retaining full copywrite ownership over the resulting song, the user would have a much stronger ownership claim over song (B).

Recent Legal Cases Shaping the Debate

Several legal challenges are shaping the boundaries of AI and copyright:

1. Thaler v. Perlmutter (2023):

Stephen Thaler sought copyright for an artwork created by an AI system called “Creativity Machine.” The Copyright Office rejected the claim, and the court upheld the decision, emphasizing that only works created by humans are eligible for copyright.

2. Getty Images v. Stability AI (Pending):

3. Class-Action Lawsuits Against OpenAI and Others:

Fair Use and Artificial Intelligence Training

Training artificial intelligence models requires vast amounts of data which are often scraped from the internet. This has raised important questions about whether using already copyrighted material in training AI models constitutes fair use—a legal doctrine which allows limited use of copyrighted works for purposes such as commentary, education, or research without needing to acquire permission from the copyright holder.

As you have probably gathered from reading, courts have not yet reached a consensus on whether using copyrighted material for training artificial intelligence models qualifies as fair use. Whether it does will hinge on several factors:

• The purpose and character of the use (sufficiently transformative or not)
• The nature of the underlying copyrighted work
• The amount used of the underlying copyrighted work
• The effect on the market for the underlying copyrighted work

As one could imagine, the stakes regarding this outcome are very high. For instance, if the courts rule that artificial intelligence training does not qualify as fair use it could disrupt the entire generative AI industry to devastating effect by frustrating the ability to train any artificial intelligence models.

What’s Next?

As artificial intelligence continues to develop and becomes more integral to creative industries, policymakers are beginning to explore how to adapt copyright law into a new paradigm. Some proposals on the horizon include:

• New copyright categories for AI-assisted works;
• Data transparency rules for training sets;
• Licensing frameworks to compensate original creators; and

Congressional hearings and studies by the U.S. Copyright Office suggest that reform is coming—but it will require a strict balancing of innovation with creators’ rights.

What This Means for Creators and Legal Professionals