Copyright Law Meets AI: Authorship, Infringement, and the Future of Creative Rights

Copyright Law in the Age of AI: Authorship, Infringement, and Creative Rights

In April 2023, a song titled “Heart on My Sleeve” went viral on music streaming platforms. It sounded like a collaboration between Drake and The Weeknd, but neither artist was involved. Instead, an anonymous producer called “Ghostwriter” used AI to mimic their voices and style. This incident highlights how AI is reshaping creative industries.

AI can now generate music, art, literature, and films that closely imitate human creativity. This challenges traditional copyright laws, which were created before machines could produce creative works. Key questions arise: Can AI-generated works be copyrighted? Is training AI on copyrighted materials infringement? How do we protect human creators while supporting technological progress?

Legal Background: Authorship and Infringement in the AI Era

The “Human Authorship” Principle

U.S. copyright law is based on the idea that only humans can be authors. The U.S. Copyright Office requires a work to originate from a human to qualify for copyright registration. Courts have upheld this principle, such as in the “monkey selfie” case where animals were ruled ineligible to hold copyrights.

This creates a problem for AI-generated works. If an AI creates a poem, painting, or song without meaningful human input, current law suggests it can’t be copyrighted. The Copyright Office confirmed in 2023 that purely AI-generated works without human involvement are not eligible for registration.

For example, Stephen Thaler tried to register an AI-created image called “Creativity Machine.” The Copyright Office rejected it, and a federal court agreed, stating only human beings can be authors. However, the UK takes a different approach, granting copyright to “computer-generated” works and assigning authorship to the person who arranged for the creation. This offers a shorter protection term but recognizes AI contributions pragmatically.

AI Training Data: Fair Use or Infringement?

AI systems learn by analyzing huge datasets, many containing copyrighted works. The big question: does training AI on these works without permission count as fair use or infringement?

AI developers argue training is transformative—like reading books to learn—citing cases where scanning books for search engines was deemed fair use. But creators disagree, especially when AI outputs compete with or replicate original works.

Visual artists and authors, including George R.R. Martin, have sued AI companies for using their works without permission. A notable case is Thomson Reuters v. Ross Intelligence, where the court ruled that using copyrighted legal databases for commercial AI training was infringement, rejecting the fair use defense.

The law here is unsettled. The U.S. Copyright Office is studying how copyright applies to AI training and outputs. Some AI firms now license data from content owners, aiming for fair compensation.

In May 2023, the Copyright Office issued guidance stating that training AI on copyrighted works can be fair use if it serves research or analytics and doesn't substitute for the original works. But unauthorized mass commercial use, especially via scraping, falls outside fair use.

Key Cases and Examples

Copyright Office Decisions on AI-Generated Works

The Copyright Office has denied copyright protection for purely AI-generated works due to lack of human authorship. Stephen Thaler’s case involving an AI-generated image was a landmark decision supporting this view.

In a more nuanced case, Kristina Kashtanova’s graphic novel “Zarya of the Dawn” had AI-generated images. The Office allowed copyright on the narrative and the curated presentation but excluded the AI-created images themselves.

Now, applicants must disclose any AI-generated content in copyright submissions. Attempting to register AI-created work as solely human-authored risks cancellation.

AI in Music: Deepfakes and Legal Responses

The “Heart on My Sleeve” AI track didn’t copy existing songs but mimicked artists’ voices and styles. Universal Music Group acted quickly to remove it, citing copyright and publicity rights. Yet, traditional copyright claims don’t fully cover this scenario.

These “deepfake” tracks raise questions about rights related to vocal timbre and likeness, which might fall under publicity rights or trademark law rather than copyright.

Tennessee’s ELVIS Act (2024) addresses this gap by protecting musicians’ voices and likenesses from unauthorized AI replication. Similar federal proposals, like the NO FAKES Act, aim to extend such protections nationwide.

Other Legal Developments

The Generative AI Copyright Disclosure Act, introduced in 2024, would require AI developers to disclose copyrighted works used for training. This transparency can help creators seek compensation.

The European Union’s AI Act and copyright directives allow text and data mining for AI research but let rights holders opt out of commercial AI training.

Litigation such as Getty Images v. Stability AI targets AI firms accused of copying millions of photos for training. Industry groups advocate for authorized and fairly compensated use of artists’ work in AI.

Legal Debates and Proposed Solutions

Should AI-Generated Works Have Copyright?

Current U.S. law excludes AI creations without human input from copyright. Supporters say copyright exists to incentivize human creativity, which AI lacks. Granting rights to AI outputs might unfairly benefit tech companies over creators.

On the other hand, denying copyright could discourage AI innovation. Solutions include requiring a minimum level of human involvement for copyright eligibility or creating a new, shorter-term legal right for AI-generated works.

For now, the focus is on clarifying when AI-assisted works qualify for copyright based on human contribution.

Clarifying Fair Use for AI Training

• Congress could create exceptions for AI training, allowing it for certain purposes while protecting creators’ markets.
• Courts may develop clearer rules through ongoing cases, distinguishing commercial from non-commercial AI uses.
• Industry licensing agreements between AI developers and content owners offer practical paths forward.

Voluntary licensing is growing, and some experts propose frameworks where creators can opt out of training or receive compensation if their work is used. The Disclosure Act supports transparency, enabling such models.

Ensuring Fair Compensation and Collaboration

To avoid disadvantaging human creators, royalty or levy systems could share revenue when AI mimics artists. Detection tools might identify AI outputs influenced by specific works.

Contractual frameworks could recognize human creators’ contributions in AI-assisted projects, ensuring credit and profit sharing. Entertainment unions have begun negotiating protections against unwanted AI exploitation.

Conclusion

AI challenges fundamental copyright principles by generating content that mimics human creativity. A balanced approach includes:

• Maintaining human authorship as a copyright foundation, with AI as a tool
• Defining clear limits for AI training under fair use
• Implementing protections like the ELVIS Act for rights beyond copyright
• Developing fair compensation models for creators impacted by AI

This is a critical moment to shape copyright laws that protect creative labor while embracing AI innovation. Thoughtful reforms can turn AI’s rise into a story of collaboration rather than conflict.

For creatives interested in learning more about AI tools and their impact on creative industries, resources like Complete AI Training offer up-to-date courses and insights.