Generative AI Copyright: Legal Foundations, Authorship, and Policy Directions for Scholars

December 18, 2025

Generative AI copyright sits at the center of today’s debate over legal protection, copyright infringement, and authorship. Generative artificial intelligence tools that create text, images, and other content from prompts, such as ChatGPT and other generative AI systems, are reshaping how we think about ownership and originality.

As new AI technologies gain ground, disputes over who, or what, owns AI-generated works challenge Copyright Office policy, copyright registration, case law, and academic theory.

Foundations of Copyright and Generative AI

Understanding copyright infringement and protection in the era of machine learning starts with traditional copyright theory and the way AI-generated material tests its boundaries. As AI developers and AI companies release increasingly powerful content-generation tools, courts and scholars must assess whether outputs from LLMs (large language models) and AI image generators qualify as protectable works. As human and machine contributions increasingly overlap, questions around originality and exclusive rights become more complex.

Historical Copyright Doctrine

Three foundational cases shape the modern view of copyright protection:

  • Feist v. Rural (1991) established that facts are not copyrightable without a “modicum of creativity.”[1]
  • Satava v. Lowry (2003) ruled that works based on natural forms require a creative contribution to receive protection.[2]
  • Early software cases, like Apple Computer, Inc. v. Franklin Computer Corp. (1983), clarified that only the expression of an idea in code, not the idea itself, is protectable.[3]

These rulings reinforce that human creativity is the cornerstone of U.S. copyright law, a principle tested by modern companies and their use of AI.

Defining AI-Generated Content

Distinguishing AI-generated content from human-directed work is central to copyright disputes involving tools like OpenAI’s ChatGPT and DALL-E, Stability AI’s Stable Diffusion, or Midjourney.

Key criteria:

  • Level of Human Input: The more creative control a user exerts, the stronger the case for authorship.
  • Role of the AI: Tools that assist, rather than autonomously generate, are more likely to support claims of human authorship.
  • Disclosure: Many publishers and institutions now require disclosure of AI’s role to ensure transparency in copyrighted materials.

The U.S. Copyright Office maintains that works created without human involvement are not copyrightable, while AI-assisted works may qualify if a human being contributed sufficiently.[4] For legal professionals and content creators, understanding the threshold between using a tool and creating a protectable work is essential.

Authorship and Ownership Challenges

The rise of generative AI systems has complicated the notion of authorship. Who owns the work: the person inputting prompts, the AI developers, or no one at all?

Human Authorship Requirement

U.S. law and the Copyright Office reaffirm that only human beings can be authors. Section 102 of the Copyright Act speaks of “original works of authorship,” and the Office’s 2025 guidance reiterates that providing a prompt alone is not enough.[5],[6] These principles have been upheld in cases like Thaler v. Perlmutter (2023), in which the D.C. Circuit Court upheld the Copyright Office’s decision that works generated solely by AI are not eligible for copyright protection because U.S. law requires a human author.[7]

Courts assess whether a human made meaningful creative decisions. If so, the resulting work may receive protection. Otherwise, machine-generated results, even if novel or visually compelling, fall outside the scope of copyright registration.

Internationally, jurisdictions like the EU also prioritize human authorship.[8] While scholars debate the viability of hybrid authorship (human and AI working together), current laws remain firmly grounded in human contribution.

Joint Authorship and Contributor Rights

Joint authorship may arise when both users and developers influence the final output.

U.S. law requires:

  • Intent to merge contributions into a single work.
  • Each contributor’s portion must be independently copyrightable.[9]

In the AI context, this is complicated. Developers shape the model, its data, and responses, while users guide specific outputs through prompts and editing. Where both parties show creative input, joint ownership could be possible, though this is still untested territory in many copyright lawsuits.

Work-for-hire doctrines also apply. If employees use generative AI technologies as part of their jobs, and human creativity is demonstrable, employers typically own the copyright.

Fair Use, Transformative Use, and Liability

The widespread use of machine learning and data mining to train AI and LLMs raises key questions about fair use, particularly for copyright owners and platforms that ingest massive datasets.

1. Purpose and Character of Use

Fair use favors transformation. AI-generated output may qualify if it adds new expression or meaning. However, outputs that closely mirror source materials (especially without visible human input) are unlikely to meet this standard.[10]

Scholars and regulators note that use of copyrighted materials in AI training datasets could be transformative if the output serves a new purpose. Still, this remains one of the most debated aspects of AI and copyright infringement law.

2. Amount and Substantiality

This factor examines how much of a copyrighted work is used. Even small excerpts, if central or recognizable, can lead to infringement claims. AI developers must ensure that outputs do not replicate signature elements of copyrighted inputs.

Many lawsuits allege that AI companies like OpenAI and Midjourney have allowed models to output text or imagery that mimics protected content. Courts will likely scrutinize how generative AI models balance creative generation with legal reuse, particularly in regards to use of copyrighted materials as training data.

3. Market Effect and Economic Harm

This factor is often the decisive one. If AI-generated content competes with or replaces original works, courts are less likely to find fair use.[11] The scale and speed at which generative AI systems can create substitutes raise real concerns for copyright holders and content creators alike.

New licensing models, watermarking systems, and litigation strategies are emerging as AI technologies become more entrenched in creative industries.

Policy, Regulation, and Future Directions

Rapid growth in AI has created an urgent need for updated policies that reflect the realities of generative AI, copyright registration, and copyright infringement.

U.S. Copyright Office Stance

The Office’s 2025 reports confirm:[11]

  • No protection for wholly AI-generated content without a human author.
  • AI-assisted works may qualify based on specific, documented human contributions.
  • Disclosure of AI involvement is mandatory during registration.

Courts have echoed these positions, noting that current law cannot accommodate entirely machine-created content. Legal battles, such as those involving LLMs trained on copyrighted datasets, continue to shape this space.

International Perspectives

Global responses vary:

  • European Union: Requires human creativity for protection and limits rights for AI-generated content.[8]
  • United Kingdom: Allows limited protection for computer-generated works, depending on who made the necessary arrangements.[12]
  • Other Countries: Most jurisdictions still hinge on the requirement of a human author, although legal interpretations vary.

This fragmented landscape creates challenges for AI companies, copyright owners, and legal professionals navigating cross-border disputes.

Research and Policy Opportunities

As generative AI tools mature, scholars and policymakers must guide doctrine. Priority areas include:

  • Attribution Models: Designing frameworks to recognize contributions from both humans and machines.
  • AI Licensing Systems: Developing fair ways to compensate original rights holders whose works are used in data mining and AI training.
  • Transparency Standards: Mandating disclosure of AI’s role in creative works to support accountability.
  • Economic Impact Studies: Analyzing how generative AI technologies influence the livelihood of content creators and creative markets.
  • Fair Use Reform: Rethinking fair use to account for transformative potential and risks from LLMs and other AI models.

Legal education programs, including WashU’s AI CLE Series, are equipping lawyers to lead in this evolving field.

Conclusion

The legal questions surrounding generative AI copyright are far-reaching. With tools like ChatGPT, Midjourney, and other AI image and text generators entering mainstream use, the need for clear guidance on authorship, ownership, and exclusive rights is more pressing than ever.

The U.S. legal system, rooted in the requirement of a human being as the author, faces pressure to adapt. Courts, lawmakers, and academics must work together to protect the rights of copyright owners and content creators, while fostering responsible innovation by AI developers and AI companies.

Active research, responsive policy, and legal reform will shape the future of intellectual property in the age of machine learning and AI technologies.

[1] Feist Publications, Inc. v. Rural Tel. Serv. Co., 499 U.S. 340 (1991).

[2] Satava v. Lowry, 323 F.3d 805 (9th Cir. 2003).

[3] Apple Computer, Inc. v. Franklin Computer Corp., 545 F. Supp. 812 (E.D. Pa. 1982).

[4] U.S. Copyright Office, Copyright and Artificial Intelligence, U.S. Copyright Office, https://www.copyright.gov/ai/.

[5] U.S. Copyright Office, Copyright Law of the United States and Related Laws Contained in Title 17 of the United States Code – October 2022 (Circular 92), (Oct. 26, 2022), https://www.copyright.gov/title17/title17.pdf.

[6] U.S. Copyright Office, NewsNet Issue 1060, https://www.copyright.gov/newsnet/2025/1060.html.

[7] Thaler v. Perlmutter, 23-5233 (D.C. Cir. Mar. 18, 2025).

[8] Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act), 2024 O.J. (L 1689) 1.

[9] 17 U.S.C. § 201 (2024).

[10] Stuart D. Levi, Mana Ghaemmaghami & MacKinzie M. Neal, Copyright Office Weighs in on AI Training and Fair Use (Skadden 2025), https://www.skadden.com/insights/publications/2025/05/copyright-office-report.

[11] United States Copyright Office, Copyright and Artificial Intelligence, Part 3: Generative AI Training Pre-Publication Version (May 9, 2025), https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-3-Generative-AI-Training-Report-Pre-Publication-Version.pdf.

[12] United Kingdom Intellectual Property Office, Copyright and AI: Consultation (2021), https://www.gov.uk/government/consultations/artificial-intelligence-and-intellectual-property-call-for-views.

Categories
Share