The rise of generative artificial intelligence (AI) has introduced dramatic shifts in creative industries. As generative AI tools become widely used in generating images, music, literature, and more, critical legal and policy issues surrounding authorship, originality, ownership, and copyrightability have emerged. There is an urgent need for an adaptation of existing legal frameworks, particularly copyright law, to address these new issues. Central to these issues is how written, audio, and visual works created by generative AI systems are treated under existing law and current Copyright Office registration practices. We discuss here the current treatment of generative AI by the Copyright Office in the context of copyright registration.
The U.S. Copyright Office (USCO) has taken a leading role in evaluating copyright and artificial intelligence through its ongoing AI initiative. In 2023, the Copyright Office issued registration guidance clarifying its position on works incorporating AI-generated content. The Copyright Office emphasized that while copyright law is technologically neutral, human authorship remains a fundamental prerequisite for protection.
According to the Copyright Office, works generated solely by AI systems, with no human contribution, are not eligible for copyright. However, works in which a human makes creative arrangements, selects or modifies AI output, or embeds AI-generated material within a larger human-generated work may qualify for registration if the human contributions are determined sufficient expressive elements to support authorship.
Under the Copyright Act and established doctrine, copyright protects "original works of authorship fixed in a tangible medium." The requirement of a human author has been emphatically affirmed in cases such as Thaler v. Perlmutter, 687 F. Supp. 3d 140, 149-50 (D.D.C. 2023). In that case, Dr. Stephen Thaler attempted to register a visual artwork titled "A Recent Entrance to Paradise," which was autonomously created by an AI system he developed called the "Creativity Machine." Thaler did not claim any human contribution to the work’s creation; rather, he listed the AI as the sole author and himself as the claimant by virtue of his ownership of the machine.
The USCO denied registration, citing the lack of human authorship. Thaler requested reconsideration twice, asserting that AI-generated works should be eligible for copyright and that ownership should vest in the machine’s owner. The USCO maintained its position that human creativity is a prerequisite to copyright, ultimately rejecting the claim. Thaler then filed suit under the Administrative Procedure Act (APA), arguing that the USCO’s decision was arbitrary and capricious.
The Court upheld the USCO’s determination. The court held that U.S. copyright law, both by its statutory language and by longstanding judicial interpretation, requires human authorship. In rejecting Thaler’s claims, the court emphasized that the Copyright Act protects "original works of authorship," and that the term "author" has always been understood to mean a human being. The court further rejected Thaler’s arguments based on ownership doctrines like work-for-hire or property law principles, concluding that these theories could not apply where no copyright existed to begin with.
The court rooted its decision in historical and constitutional policy rationales. It noted that the Framers of the Constitution granted Congress authority over copyright to promote the progress of science and the useful arts by incentivizing human creativity. The incentive structure underpinning copyright—granting exclusive rights to spur human expression—has no application to machines or non-human actors. Because AI systems do not require incentive to create and cannot be deterred or encouraged by rights and remedies, the fundamental purpose of copyright protection is not served by extending authorship to AI.
The Thaler decision affirms a bedrock principle of U.S. copyright law: the protection of works under copyright depends upon human creativity. This requirement acts as a gatekeeper against efforts to extend intellectual property rights to machine-generated content in the absence of meaningful human input. While the court acknowledged the potential complexities that may arise as AI becomes more deeply integrated into creative processes, it held that those questions were not at issue in the case before it—where no human had contributed at all.
The Copyright Office’s registration guidance aligns with this precedent by requiring applicants to disclose and disclaim AI-generated portions of works and to focus registration claims solely on the human-authored components. Prompts alone, no matter how complex or refined, are insufficient to establish authorship, as they lack the degree of expressive control required by law. The ruling in Thaler reinforces the Office’s position and provides a judicial foundation for its ongoing application of the human authorship requirement.
The Copyright Office’s Copyright and Artificial Intelligence, Part 2: Copyrightability report outlines specific criteria used to evaluate whether AI-generated content is part of a protectable work. It examines whether a human made meaningful creative decisions reflected in the final product. This includes scenarios where a human arranges AI-generated content, modifies outputs creatively, or combines them with original expressive elements.
In assessing authorship, the Copyright Office conducts a fact-specific inquiry that centers on whether the human claimant exercised control over the expressive elements of the work. The mere triggering of an AI system to generate content—by providing prompts or instructions—is generally insufficient. Instead, the Office seeks evidence that the claimant contributed original expression by selecting, coordinating, or arranging AI-generated content in a creative way or by meaningfully modifying the outputs to shape the final expression.
The Copyright Office has reiterated that generative AI tools, like other technological aids, can support but not supplant human authorship. The difference lies in whether the human uses the AI as a tool to express original ideas, or whether the AI system independently determines the expressive elements. The use of AI in assistive or augmentative capacities does not preclude copyrightability, but AI-generated outputs must be embedded in a larger human-generated work or reflect a human author’s creative control.
The Office’s inquiries into whether there is sufficient human authorship may include questions such as:
For example, if a designer uses an AI program to generate dozens of visual motifs and then selects a handful, modifying their color, composition, and integration into a human-designed layout or publication, the resulting work may contain protectable human-authored elements. By contrast, if a user generates a single AI image using a descriptive prompt and submits it without further creative refinement, the output would not meet the threshold of human authorship.
Another illustrative scenario involves a comic book creator who uses AI to generate background art, which is then incorporated into a graphic novel that features original characters, plot, and dialogue written by the author. If the AI backgrounds are creatively chosen, altered, and integrated into the panel layout, and if the overall selection and coordination reflect human authorship, the work may be registrable with a disclaimer as to the AI-generated elements.
Copyright holders must clearly identify and disclaim any AI-generated portions that do not reflect human authorship in their registration applications. Applicants are expected to describe the human contributions in detail—for instance, by identifying which parts of a visual work were modified by the claimant and how. The Copyright Office may register the human-created parts of a work while excluding purely AI-generated material, and failure to disclose the use of generative AI may result in cancellation of a registration.
This fact-driven, case-by-case analysis ensures that copyright protection remains grounded in the core principle of human creativity, while providing a flexible framework for creators who use generative AI tools in an assistive capacity.
In copyright registration practice, the Copyright Office requires applicants to disclose the use of generative AI systems and clearly differentiate between human and machine-generated content. This identification and disclaimer requirement stems from the Office’s March 2023 policy guidance and reflects the fundamental principle that copyright protects only human authorship.
When a work submitted for registration contains AI-generated material, the applicant must identify such content in the application and disclaim authorship of those specific portions. This is typically done by providing a statement in the “Author Created” and “Limitation of Claim” sections of the application, explaining what was created by the human author and what, if any, was generated by an AI system. The Office provides examples, such as: “AI-generated text excluded; human-authored text claimed” or “Image generated by Midjourney AI excluded; layout and captions claimed.”
Failure to properly identify and disclaim AI-generated content may result in cancellation of the registration or refusal to register the claim. The Office treats nondisclosure of AI-generated elements as a material misrepresentation, which can jeopardize the validity of the registration and undermine the applicant’s ability to enforce rights in court.
The degree of human authorship is critical. If a work consists primarily or entirely of AI-generated content without sufficient human involvement, the Office will reject the application altogether. However, where the human author contributes creatively—for example, by arranging AI-generated elements, modifying them meaningfully, or incorporating them into a larger original work—registration may be granted for the human-authored portions.
The Copyright Office has further clarified these procedures through updates to the Compendium of U.S. Copyright Office Practices and has held public webinars to assist applicants in navigating the registration process. These efforts form part of the Office’s broader copyright office’s AI initiative and reflect a commitment to clarity and transparency as creators adapt to new tools.
In sum, the registration framework requires a candid accounting of AI involvement. This approach both preserves the integrity of copyright’s human authorship requirement and facilitates the responsible integration of generative AI into creative practice. As generative AI technology evolves, applicants must remain attentive to changes in the law and USCO's guidance to ensure their works are compliant with the requirements of copyright law and USCO practice.
The proliferation of generative AI models, such as Stable Diffusion, DALL-E, and GPT-based systems, has led to a wave of copyright litigation and regulatory scrutiny. AI developers now face legal challenges involving claims of unauthorized use of copyrighted content as AI training data and in AI generative output. These legal issues cut to the core of what constitutes copyright infringement in the age of machine learning and content synthesis.
At the heart of these disputes is the training process itself. Generative AI systems are typically trained on massive datasets that often include copyrighted material scraped from the internet, including news articles, books, artwork, and photographs. Plaintiffs have asserted that using these materials without authorization constitutes unlawful reproduction and distribution under the Copyright Act giving rise to infringement claims. Defendants, in contrast, argue that the ingestion of content into training models is transformative and constitutes fair use, especially when the AI does not retain or replicate specific copyrighted elements in a recognizable form.
In one of the most closely watched cases, The New York Times filed suit against OpenAI and Microsoft, alleging that millions of its articles were used without permission to train large language models, including ChatGPT. The Times contends that this practice not only infringes its copyrights but also threatens its business model, as the AI system can generate summaries or reproductions of paywalled content that compete directly with the original. The lawsuit underscores questions about whether AI output that mimics or paraphrases source material constitutes derivative works under copyright law.
Similarly, Getty Images has filed litigation against Stability AI in both the United States and the United Kingdom, alleging that Stability used Getty’s licensed images without consent to train the Stable Diffusion image generation model. Getty asserts that Stable Diffusion outputs images that are substantially similar to or directly derived from Getty’s copyrighted content, and in some cases even replicate its watermark. Getty’s claims focus not only on infringement through training but also on the risk of output-based infringement and reputational harm due to association with unlicensed or low-quality derivatives.
These lawsuits raise complex and unresolved legal questions. Among them: Is training on copyrighted works without permission a violation of the reproduction right? Do AI-generated outputs constitute derivative works? What role does the fair use doctrine play in this context? And what obligations do AI companies have in disclosing or licensing their training data?
While the Copyright Office has not yet adopted a formal position on the legality of training AI models using copyrighted material, it has acknowledged in its reports that this issue presents serious licensing considerations. The Office has also noted that fair use may serve as a potential defense, though its application is highly fact-specific and uncertain in this novel context. Some advocates argue that AI training for scientific research or non-commercial uses may be defensible under fair use, while others contend that commercial AI development based on unlicensed AI training data erodes the rights of content creators.
As these cases progress through the courts, they are likely to shape the contours of liability for AI developers and clarify how existing law applies to training AI, generating outputs, and distributing derivative works. In the meantime, they highlight the growing tension between innovation in generative AI technology and the foundational rights of copyright holders.
Congress, the Copyright Office, and various stakeholders continue to debate broader legislative responses to the rise of generative AI systems. Legislative proposals such as the No FAKES Act and the Generative AI Copyright Disclosure Act reflect an increasing willingness to regulate AI-generated content. These measures focus on key concerns such as unauthorized digital replicas, the protection of personal likeness and voice, and transparency through mandatory disclosure of AI-generated content.
The Copyright Office’s public listening sessions and its Notice of Inquiry have revealed widespread concern among creators, copyright holders, and AI developers. Participants have voiced apprehension about copyright infringement, the erosion of licensing markets, and the displacement of human authorship in sectors such as journalism, illustration, and music. Many emphasize the need to further the constitutional goals of promoting the progress of science and the useful arts, while also ensuring that legal frameworks are adaptable to technological evolution.
Internationally, the approach to AI-generated content varies significantly, underscoring the growing need for global harmonization. The United States maintains a strict human authorship requirement, and the Copyright Office has consistently stated that AI-generated material lacking sufficient human creative input is not eligible for protection under the Copyright Act. However, other major jurisdictions have adopted different policies.
For example, the United Kingdom recognizes copyright in computer-generated works under Section 9(3) of its Copyright, Designs and Patents Act 1988, which provides that for works generated by a computer in circumstances such that there is no human author, the author is deemed to be the person “by whom the arrangements necessary for the creation of the work are undertaken.” This approach effectively provides copyright protection for some AI-generated works, although the scope and enforceability of these rights remain debated.
In contrast, the European Union has taken a more cautious and structured approach. The EU AI Act and other pending legislation do not currently extend copyright protection to AI-generated content, but the European Parliament has considered proposals for sui generis rights or other regulatory mechanisms to manage AI-generated outputs. Additionally, the EU Copyright Directive gives authors and publishers the right to opt out of having their works used for text and data mining, which impacts the permissibility of using copyrighted materials as AI training data.
These divergent approaches demonstrate the difficulty of these issues and the need for action in determining solutions through national copyright laws and innovation policies. As AI-generated content becomes more prevalent in cross-border creative markets, inconsistencies in protection and enforcement could create friction. This makes international dialogue and technology law review essential to harmonize standards, clarify authorship, and balance the interests of innovation and intellectual property protection.
Generative AI presents a rapidly evolving challenge to traditional copyright doctrines. The legal implications of AI-generated material require ongoing adaptation of copyright law and policy. While current registration guidance confirms that works generated solely by AI systems are not protected, works incorporating AI-generated content may qualify if a human being contributes sufficient expressive elements.
As AI developers, creators, and policymakers grapple with the legal implications of artificial intelligence, the importance of maintaining a clear boundary between human authorship and AI output is critical. The copyright office continues to refine its policies, monitor developments, and participate in ongoing public inquiry seeking input on the treatment of AI-generated works. Through this process, the law will hopefully find an equitable balance between creative expression, technological development, and the rights of copyright holders.
© 2025 Sierra IP Law, PC. The information provided herein does not constitute legal advice, but merely conveys general information that may be beneficial to the public, and should not be viewed as a substitute for legal consultation in a particular case.
"Mark and William are stellar in the capabilities, work ethic, character, knowledge, responsiveness, and quality of work. Hubby and I are incredibly grateful for them as they've done a phenomenal job working tirelessly over a time span of at least five years on a series of patents for hubby. Grateful that Fresno has such amazing patent attorneys! They're second to none and they never disappoint. Thank you, Mark, William, and your entire team!!"
Linda Guzman
Sierra IP Law, PC - Patents, Trademarks & Copyrights
FRESNO
7030 N. Fruit Ave.
Suite 110
Fresno, CA 93711
(559) 436-3800 | phone
BAKERSFIELD
1925 G. Street
Bakersfield, CA 93301
(661) 200-7724 | phone
SAN LUIS OBISPO
956 Walnut Street, 2nd Floor
San Luis Obispo, CA 93401
(805) 275-0943 | phone
SACRAMENTO
180 Promenade Circle, Suite 300
Sacramento, CA 95834
(916) 209-8525 | phone
MODESTO
1300 10th St., Suite F.
Modesto, CA 95345
(209) 286-0069 | phone
SANTA BARBARA
414 Olive Street
Santa Barbara, CA 93101
(805) 275-0943 | phone
SAN MATEO
1650 Borel Place, Suite 216
San Mateo, CA, CA 94402
(650) 398-1644. | phone
STOCKTON
110 N. San Joaquin St., 2nd Floor
Stockton, CA 95202
(209) 286-0069 | phone
PORTLAND
425 NW 10th Ave., Suite 200
Portland, OR 97209
(503) 343-9983 | phone
TACOMA
1201 Pacific Avenue, Suite 600
Tacoma, WA 98402
(253) 345-1545 | phone
KENNEWICK
1030 N Center Pkwy Suite N196
Kennewick, WA 99336
(509) 255-3442 | phone
2023 Sierra IP Law, PC - Patents, Trademarks & Copyrights - All Rights Reserved - Sitemap Privacy Lawyer Fresno, CA - Trademark Lawyer Modesto CA - Patent Lawyer Bakersfield, CA - Trademark Lawyer Bakersfield, CA - Patent Lawyer San Luis Obispo, CA - Trademark Lawyer San Luis Obispo, CA - Trademark Infringement Lawyer Tacoma WA - Internet Lawyer Bakersfield, CA - Trademark Lawyer Sacramento, CA - Patent Lawyer Sacramento, CA - Trademark Infringement Lawyer Sacrament CA - Patent Lawyer Tacoma WA - Intellectual Property Lawyer Tacoma WA - Trademark lawyer Tacoma WA - Portland Patent Attorney - Santa Barbara Patent Attorney - Santa Barbara Trademark Attorney