AI, Authorship, and Copyright: A Comparison Between the United States and the European Union

By: Addison Hichman

As artificial intelligence (AI) increasingly changes the way art, music, and literature are created, the copyrightability of these creative works has become an important consideration. Courts and policymakers are now forced to ask: Who is the author of an AI-generated work? Copyright protection has long been tied to human creativity, but the rise of autonomous machine outputs challenges this basic premise. This article compares how the United States (U.S.) and the European Union (EU) have addressed the issue of copyright authorship in AI-generated works, focusing on the recent U.S. Thaler v. Perlmutter decision and the U.S. Copyright Office’s Copyright and Artificial Intelligence, Part 2: Copyrightability Report in comparison to the EU’s Artificial Intelligence Act and Code of Practice.

Copyright protection in both jurisdictions is grounded in human authorship. In the U.S., the Copyright Act of 1976secures rights to “original works of authorship,” a phrase consistently interpreted to require a human creator. This principle flows from Article I, § 8 of the Constitution and the U.S. Copyright Office’s (USCO) Compendium of Copyright Office Practices, which reinforces the longstanding view that copyright requires human authorshipThaler v. Perlmutter, decided on March 18, 2025, reaffirmed this principle. In the EU, the Court of Justice of the EU requires that a work reflect the “author’s own intellectual creation,” a standard reaffirmed in Directive 2019/790/EU. These baselines mean that works created entirely by machines are excluded from protection, though both jurisdictions are still grappling with authorship when human and machine contributions are intertwined.

In the U.S., two main 2025 developments influenced the ability to receive copyright protection for AI-generated works. The first was the D.C. Court of Appeals decision in Thaler. Dr. Stephen Thaler attempted to register A Recent Entrance to Paradise, a work produced solely by his AI system, the Creativity Machine. He listed the AI as the author and himself as the owner. The USCO denied the application, and the court affirmed, holding that the Copyright Act requires authorship by a human being. The court emphasized that its role was to apply the statute as written and that any expansion of authorship to include AI must come from Congress. It declined to address scenarios where humans contribute to the output, leaving the status of hybrid works unresolved.

The second 2025 development was the USCOCopyright and Artificial Intelligence, Part 2: Copyrightability Report, released on January 29, 2025. The report provided: 

“The Office agrees that there is an important distinction between using AI as a tool to assist in the creation of works and using AI as a stand-in for human creativity. While assistive uses that enhance human expression do not limit copyright protection, uses where an AI system makes expressive choices require further analysis. This distinction depends on how the system is being used, not on its inherent characteristics.” 

These acts of modification or arrangement qualify as authorship because they introduce expressive elements attributable to the human author rather than the machine. The report also addressed what it called “expressive inputs.” If a human exerts creative control over the process—such as by drafting original textual material that guides the AI or embedding unique artistic choices in the prompts—then the resulting work may reflect the author’s own intellectual contribution. However, merely supplying generic prompts, without more, does not involve sufficient human expression.

Finally, the report clarified that the inclusion of AI-generated material within a larger human-authored work does not bar copyright protection for the human-authored portions. In such cases, protection extends only to the elements contributed by the human creator, not to the solely AI-generated content itself. After an extensive review of public comments, the USCO concluded that existing copyright law is flexible enough to address the challenges posed by AI and that new legislation is not presently required. Together, the Thaler decision and the USCO’s guidance reinforce a clear principle: copyright remains tied to human creativity, but there is space for protection when human authorship is perceptible in AI-assisted works.

The EU has taken a more comprehensive regulatory approach. The Artificial Intelligence Act, adopted in June 2024, regulates AI systems in the EU and classifies them into risk categories ranging from unacceptable to low risk. While initially not focused on copyright, the rise of generative AI prompted drafters to include provisions specific to general-purpose AI (GPAI) models. These provisions require transparency regarding training data, technical documentation, and policies to respect EU copyright law. Providers of GPAI models must publicly disclose a sufficiently detailed summary of the content used for training, including copyrighted works, while balancing trade secrets and confidentiality concerns.

The European Commission’s General-Purpose AI Code of Practice, released in July 2025, supplements the AI Actby requiring developers to disclose data sources, ensure lawful access to training materials, and prevent outputs from reproducing protected works. Analysts note that although these copyright provisions are relatively mild, they impose compliance obligations on developers and provide rightsholders with new tools to uncover potential infringement. The EU Copyright Directive (Directive 2019/790/EU) further interacts with the AI Act by granting rightsholders the ability to opt out of text and data mining, which raises compliance costs and limits the availability of lawful training data

Beyond the AI Act, the baseline requirement of authorship in EU copyright law continues to be grounded in the jurisprudence of the Court of Justice of the European Union (CJEU), which requires works to reflect the “author’s own intellectual creation.” The DSM Directive does not define authorship in relation to AI-generated content and does not introduce provisions recognizing AI as a creator, reflecting the broader international uncertainty over the status of AI-generated works. The CJEU confirmed in cases such as Infopaq International A/S v. Danske Dagblades Forening (C-5/08) and Eva-Maria Painer v. Standard VerlagsGmbH that copyright requires the expression of an intellectual creation reflecting the author’s personality—criteria that cannot be satisfied without human input. Consequently, works generated fully autonomously by AI are excluded from copyright protection under current EU law.

The DSM Directive nonetheless suggests how AI-assisted works could receive protection if a human author maintains creative control over the process. In this scenario, the AI system functions like a camera or computer, serving as a tool that facilitates human expression rather than replacing it. The question of ownership remains unsettled, and EU law provides that copyright initially vests in the author, meaning the human operator who provides creative input would likely be considered the owner in AI-assisted works. However, when AI operates independently, no claimant is apparent because EU law does not recognize non-human entities as authors or rights holders. Proposals to vest ownership in the developer or owner of the AI system remain controversial, as current law does not treat the mere provision of technology as sufficient to establish authorship.

A comparison of the two jurisdictions shows that both the U.S. and the EU continue to insist that copyright protection requires human authorship. In the U.S., this principle has been reaffirmed in Thaler and the U.S. Copyright Office’s Copyright and Artificial Intelligence, Part 2: Copyrightability Report, which confirmed that copyright protection attaches only where a human contributes original expression. In the EU, the Court of Justice of the EU has long required that works reflect the “author’s own intellectual creation,” a standard carried forward in Directive 2019/790/EU and emphasized in commentary that fully autonomous AI outputs do not satisfy this test. The AI Act and its July 2025 Code of Practice reinforce this baseline by imposing transparency obligations on developers of general-purpose AI models and requiring them to respect copyright law.

However, the U.S. approach is judicial and administrative, focusing on questions of copyrightability under existing statutory language, with no present effort to alter the framework. By contrast, the EU integrates copyright questions into a broader regulatory scheme, embedding them within a comprehensive system of AI governance that requires disclosures about training data and imposes compliance obligations on developers.

At the same time, both jurisdictions acknowledge the possibility of protection for AI-assisted works where humans exercise meaningful creative control. The U.S. Copyright Office has made clear that modification, arrangement, or other human choices may qualify as authorship, while EU commentary similarly treats AI as a tool comparable to a camera or computer, allowing protection when originality can be traced to human input. Yet both systems leave the precise threshold of necessary human contribution undetermined.

AI’s growing role in creative industries highlights how essential copyright remains for protecting artistic expression and for enabling creators and companies to secure the economic and reputational value of their brands. Both the U.S. and the EU maintain that AI itself cannot be the author, but each recognizes that when human creators use AI as a tool and contribute original expression, their works may still qualify for protection. The precise threshold of human input sufficient to establish authorship remains unsettled, leaving a gray area that demands close attention. Creators must continue to monitor these developments as the limits of copyright authorship are tested.

Leave a Reply

Your email address will not be published. Required fields are marked *