Authors: Juan Vasquez and Rick Sanchez
Post Date: February 12, 2025
Introduction
Artificial intelligence (AI) tools have become increasingly capable of generating creative content, from vivid artwork and music to literary text, raising fundamental questions about copyright law. Can something created by a machine be protected by copyright, and if so, who is the author? To address these issues, the U.S. Copyright Office released a major report on January 28, 2025, examining the copyrightability of AI-generated content. This report, the second part of a series on Copyright and AI, provides guidance on how existing copyright principles apply when AI is used to create works, and it offers recommendations for navigating authorship and ownership in this new frontier.
For intellectual property lawyers, the report’s findings are essential reading: they clarify how the Copyright Office will treat AI-generated works, the legal implications for registration and ownership, and what policy challenges lie ahead. This article distills the report’s key points and analyzes their implications.
Summary of Key Points from the Copyright Office’s AI Report
The Copyright Office’s January 2025 report emphasizes that existing copyright law is generally up to the task of handling AI-generated content. In other words, there is no immediate need for a special new “AI copyright law.” Longstanding doctrines, especially the requirement of human authorship, remain the touchstone for deciding whether AI-involved works can be protected. Below are the report’s central conclusions regarding AI-generated material and copyrightability.
Human Authorship is Essential
The report reaffirms the foundational principle that copyright protects only the original expression of human authors. Using an AI as a tool to assist human creativity does not, by itself, undermine the availability of copyright protection. If a person uses AI software in the creative process, much like using a camera, a paintbrush, or photo-editing software, and retains creative control over the work, the resulting output can still qualify for protection as that person’s creation. The Copyright Office gave examples of “assistive” uses of AI that do not deprive a work of protection: for instance, an artist employing AI for minor enhancements (like color correction or sharpening an image), or a writer using a generative text tool to brainstorm ideas or produce an outline that the writer then develops in their own words. In such cases, the human is augmented by AI, but not replaced, the human’s creative choices are what give rise to authorship.
Pure AI Creations Are Not Copyrightable
In contrast, the Office draws a firm line when AI is used as a stand-in for human creativity, meaning the AI generates the expressive content with minimal or no creative input from a human. The report concludes that works produced entirely by an AI without sufficient human control over the expressive elements cannot be protected by copyright. Any material that is “purely AI-generated,” in the Office’s words, falls outside the scope of copyright because it lacks a human author.
This aligns with a growing body of case law and precedent holding that non-humans cannot be authors under U.S. law. The Copyright Office’s own recent experiences reinforce this principle: it has denied or canceled registrations for AI-generated works where it determined that a human did not actually create the expressive content. One high-profile example was the graphic novel Zarya of the Dawn, for which the Office initially issued a registration but later revoked protection for the AI-generated images after discovering they were produced by the Midjourney AI system. In that matter, the Office reasoned that while the human artist had guided the AI with text prompts and made some selections, the actual expressive details of the images were generated by the machine, not the artist, therefore the artist could not be considered the “author” of those visual elements.
Text Prompts Alone Don’t Qualify as Authorship
A particularly significant point in the report is the treatment of text prompts, the written instructions a user gives to a generative AI (like “create a painting of a sunset over the mountains in the style of Van Gogh”). The Copyright Office concluded that entering prompts, by itself, is generally insufficient to make the user the author of the AI’s output. This is because current generative AI technology operates in a largely autonomous way: the user’s prompt might set a general direction or idea, but the AI is doing the heavy lifting of “filling in” creative details and often producing unpredictable results.
The Office’s view is that a simple prompt is more like providing an idea or suggestion, which is not protectable expression, rather than executing creative control over the final expression. The report supports this stance by noting practical realities: identical prompts can yield different outputs, and users typically accept one of the AI-generated results rather than determine every element of what the output looks or sounds like. Even iteratively refining prompts (entering a prompt, getting an output, then tweaking the prompt and trying again) does not change the analysis, according to the Office. No matter how many rounds of prompts are done, ultimately “the final output reflects the user’s acceptance of the AI system’s interpretation, rather than [the user’s] authorship of the expression it contains.”
In short, at the current stage of technology, the human user’s role in a prompt-based generation process is too far removed from the expressive decisions made in the output. The Office does leave the door open to a different conclusion in the future, if AI tools evolve to allow users much more direct and predictable control over the content, then perhaps a prompt could embody sufficient human creativity to confer authorship. But as of now, simply typing a description and letting the AI do the rest will not earn the user a copyright in the result. (Notably, the report adds that a very detailed and creative prompt might itself be protected as a literary work, the text of the prompt is authored by a human, but that does not extend to making the AI-generated output the human’s creation.)
Protecting Human Contributions in AI-Assisted Works
While purely machine-made expression isn’t copyrightable, the report makes clear that not all works involving AI are automatically disqualified. The key is whether there is identifiable human-authored expression in the work. If a work contains a mix of human and AI content, any original material contributed by the human is still eligible for copyright protection.
The Office emphasizes that a human author does not lose rights simply because they used an AI tool as part of their process. For example, if a photographer takes a picture and then uses an AI-powered filter or editing tool to enhance it, the final image is still protected by copyright as the photographer’s work, since the essential creative elements (the composition, the subject choice, etc.) came from the photographer. Similarly, if a novelist writes a draft and then asks ChatGPT or a similar AI to rewrite portions or suggest alternate phrasings which the novelist selectively adopts, the human-authored narrative remains protected, the AI’s assistance doesn’t negate the author’s ownership of the text that the author actually crafted or selected.
The report goes further in describing scenarios where part of an AI-generated output may be deemed copyrightable due to human input. One scenario involves “expressive inputs” provided by a human to the AI. An example is a hand-drawn sketch that a human feeds into an image-generating AI, asking the AI to produce a refined, color illustration based on that sketch. In such a case, the human-created sketch is an expressive input that significantly shapes the final output. The Copyright Office suggests that if the original human expression remains perceptible in the AI-generated output, then that portion of the output derives from human authorship and can be protected. In effect, the human author of the sketch would still be the author of those aspects of the AI-assisted image that reproduce the creativity of the sketch.
Another scenario involves creative modifications or compilations made by a human using AI outputs as raw material. The report reaffirms that if a person takes AI-generated content and then exercises creative judgment in selecting, arranging, or editing it, the result may cross the threshold of originality as a human-authored work. In other words, someone can curate or modify AI outputs in an original way and still receive copyright for their contributions.
The Copyright Office also considered situations where a registrant provided a copyrighted illustration as input to an AI and got a modified image; in such cases, the Office found the human remained the author of the unaltered or recognizable portions of their original art in the new image, though the registration would explicitly exclude the AI-generated modifications. Similarly, if a graphic designer uses fragments of several AI-generated images and combines them into a collage with a particular aesthetic layout, that arrangement (if it reflects creative choices beyond a trivial level) could be a protectable work of authorship owned by the designer.
The crucial point is that any expressive choices made by a human in the process can give rise to copyright, even if the ingredients include AI-produced material. The Office does caution that not every minor edit or arrangement will qualify, the human contribution must still meet the low bar of “minimum originality” under the law. Simply rearranging two or three AI-generated elements might be too trivial to count, but a more complex or nuanced arrangement likely would pass muster. And as always, these questions will be decided case by case, looking at the specific facts of how the work was created and what the human did.
No New Special Protection for AI Outputs (For Now)
Finally, after surveying the landscape, the Copyright Office concluded that additional legislation or sui generis rights for AI-generated works are not warranted at this time. Some stakeholders had urged creating new rules to expressly grant protection to AI-generated content or to the companies developing AI, citing reasons like encouraging innovation or keeping pace with other countries. The Office, however, was not convinced. It reasoned that creators of AI technology already have plenty of incentives (such as market rewards, trade secrets, patents for the software, etc.) to continue innovating without needing copyright ownership of the outputs. Moreover, granting copyright to machine-generated works could pose risks to the creative ecosystem. In particular, the Office expressed concern that a flood of AI-generated works with copyright might “crowd out or devalue human-created content” in the marketplace.
This reflects an underlying policy judgment that human creativity should be incentivized and rewarded and that giving equal legal status to automated creativity might undermine those incentives. The report also considered the international context and found that the emerging approaches in other jurisdictions generally align with the U.S. stance, so the fear of falling behind globally did not justify a policy shift.
In sum, the Office’s message was to hold the line on traditional principles: as long as a work has human creative input, it can be protected, and if it doesn’t, it won’t receive copyright, no new rights or legal carve-outs for AI are deemed necessary. That said, the Office will continue to monitor technological and legal developments closely, and it plans to issue further guidance (including updates to the Compendium of U.S. Copyright Office Practices) to help the public and practitioners apply these principles in concrete cases.
Legal Implications for Copyright Registration and Ownership
The Copyright Office’s reaffirmation of human authorship as a prerequisite has immediate practical implications for how works involving AI are registered and how ownership rights are determined. One clear takeaway is that when filing for copyright registration, applicants must be transparent about any AI-generated material in the work and cannot claim authorship over those portions. In fact, the Office’s policies (including a March 2023 registration guidance) require that authors disclose the use of AI content in their works and exclude the non-human contributions from their copyright claim.
This means if a novelist uses an AI to generate some paragraphs in a book, or an artist incorporates an AI-created background in a painting, they should identify those portions in the application and make clear they are not claiming them as their own expression. The reason is straightforward: including AI-generated material without disclosure could result in an invalid or fraudulently obtained registration, since the certificate would imply the entire work is human-created when it is not.
The Zarya of the Dawn incident serves as a cautionary tale, the author’s initial copyright registration did not mention that the images were AI-produced, and once the Office learned of it, they moved to cancel the registration as to those images for being “inaccurate or incomplete” in claiming authorship. To avoid such outcomes, attorneys should advise clients to be forthright and precise in applications, listing any AI-generated content as unprotectable material. Indeed, the Copyright Office now scrutinizes works suspected of having AI elements, and examiners may inquire further or require a disclaimer for the AI portions before granting registration.
Because copyright protection will only extend to the human-created elements of a work, the registration process essentially carves out what the applicant truly authored. For example, if a client designs a poster and used an AI to generate the background imagery, the registration might be issued for “text and graphic elements excluding AI-generated background art.” The certificate would thereby protect the aspects that the human designer created (say, the text layout, any original graphics, or the selection and arrangement of elements) and not the AI-produced background. Importantly, even if the Office registers a mixed work, the AI-generated parts remain in the public domain by law, owned by no one. The registrant holds a copyright only in the original contributions they added.
From an ownership perspective, this means that no one can claim exclusive rights over the AI-generated portions, not the user who prompted the AI, not the developer of the AI, and certainly not the AI itself. Those portions are free for others to use, copy, or build upon without infringing copyright (provided they don’t inadvertently copy the human-authored parts that are protected).
This division of protected vs. unprotected content has several legal ramifications. First, it affects enforcement: If a third party copies an AI-generated image that your client included in their work, your client may have no grounds to sue for copyright infringement over that image, because the image itself wasn’t protected in the first place. However, if the third party also copied the way your client arranged that image alongside other elements or copied accompanying text that was human-written, those acts could still infringe the human-authored components.
Second, the need to identify AI content puts a premium on documenting the creative process. Lawyers should counsel creators to keep records of their use of AI (prompts given, drafts generated, edits made, sources of any inputs, etc.) so they can later demonstrate what contributions were human vs. machine. Not only will this help in preparing accurate registrations, but it could be vital if a registration is challenged or if another party claims the work isn’t original.
Third, companies and authors might need to adjust their IP strategies. If a business is generating large volumes of AI-created content (say, marketing images or blog posts), they must realize that much of that content may be uncopyrightable under this framework. They cannot rely on copyright to stop competitors from reusing purely AI-generated material.
Other legal tools would have to fill that gap, if possible, such as trademark law to protect a brand logo that was AI-designed (trademark law, unlike copyright, does not require human authorship for a logo, though it requires distinctiveness), or using contracts and confidentiality to protect AI outputs that are kept as trade secrets.
Conversely, businesses may decide to involve human creators more directly in the process (for example, having a human significantly edit or curate AI outputs) to ensure the final product qualifies for copyright. The Copyright Office’s report effectively puts creators on notice: if you want the exclusive rights and remedies that copyright provides, you need to inject a human touch into the work’s creation and be able to prove it.
Finally, ownership questions might arise in collaborative settings involving AI. If multiple people contribute human authorship to a work that also includes AI material, general joint authorship principles would apply among the people (assuming each intended to merge their contributions into a unitary work). But none of them can claim the AI’s portion as a joint author, since the AI isn’t a legal author at all. This is an odd situation where a collaborative work could have gaps of public-domain content interspersed with co-owned content. For instance, two graphic designers might jointly create a collage, part of which uses AI-generated imagery; they could be co-authors of the overall collage, but the AI images within it are not owned by either of them (yet their particular cropping or arrangement of those images is protected). Such complexities mean IP lawyers should pay close attention to how contracts and licenses are worded. If licensing a work that contains AI material, the license should clarify which elements are actually protected and being licensed, and perhaps disclaim any rights in elements that are unprotectable. Likewise, if commissioning a work, a client may want warranties that either no AI was used without disclosure or, if it was, that the human creator has added sufficient original expression to secure copyright in the end product.
Policy Considerations and Challenges for Copyright Law
The January 2025 report not only lays down rules for what is or isn’t protectable, it also grapples with broader policy questions and acknowledges the challenges that AI-generated content poses to the existing copyright framework. One major consideration is whether the current law’s human-authorship requirement will continue to suffice as AI technology advances.
The Copyright Office’s stance for now is that current law is “adequate and appropriate,” and it emphasizes that copyright has historically adapted to new technologies on a case-by-case basis. Essentially, rather than preemptively rewriting the rules for AI, the Office believes we can apply core principles (like originality and human creativity) to new scenarios as they arise. This approach has the benefit of flexibility, courts and the Office can evaluate each AI-involved work on its facts, but it also means some uncertainty will persist. Creators may not always know in advance how much human input is “enough” to qualify for protection, especially as AI systems become more sophisticated creative partners. The case-by-case mantra implies that clear lines are hard to draw in statute right now, and policy will evolve with real-world test cases.
Another challenge highlighted is the difficulty of distinguishing AI-generated content and assessing human contribution. AI outputs don’t come with a watermark indicating how much is machine-made. As AI becomes better at mimicking human style, it might be hard even for experts (let alone the Copyright Office) to tell if a work was AI-generated or just inspired by AI or entirely human. This raises enforcement and administrability issues. The Copyright Office will have to rely largely on self-reporting by applicants to flag AI content, a system that might be gamed by those who hope to sneak through registrations for AI works. The Office’s commitment to update its Compendium and provide guidance suggests it’s aware of these pitfalls and is trying to create standardized practices (for example, possibly requiring a statement of AI use in applications). Policy-wise, there could be discussions in the future about penalties for failing to disclose AI involvement or even technical measures to detect AI-generated material in submissions. For now, the onus is on creators and their counsel to be transparent.
The report also wades into the debate on incentives and the impact on human creativity. By denying copyright to pure AI works, the policy effectively pushes authors toward maintaining a significant role in the creative process. This could be seen as preserving a kind of “human quota” in creative output. Some technologists argue that this might discourage innovation or the use of helpful AI tools. For example, why penalize a creator for using a tool that perhaps generates a whole image from a prompt if the end result is valuable? The Copyright Office’s answer is that the value to society of human creativity is paramount, and that AI can still be used in supportive ways without displacing the human’s role. In fact, the Office explicitly differentiates between AI as an assistive tool versus AI as an autonomous creator. This policy perspective resonates with the constitutional purpose of copyright “to promote the progress of science and useful arts,” which, historically interpreted, means incentivizing human authors to produce creative works. If AI could generate millions of artworks at the click of a button and each were copyrighted, it might lead to an abundance of protected works with minimal creative labor, potentially clogging the system and diminishing the relative value of works produced by humans. The report cites concerns that according full copyright to AI outputs could devalue human-created content in the market, suggesting a cultural policy choice to privilege the human element.
There are also competitive and economic policy angles. Some stakeholders worry that if the U.S. is too strict on refusing protection to AI-assisted creations, it could lag behind other countries that might grant such protection (impacting industries like entertainment or software where AI is used). However, the Copyright Office observed that most other countries are grappling with the same issue and many are arriving at a similar conclusion requiring human creativity. This reduces the risk of the U.S. being an outlier. In fact, the Office dismissed the notion that international competition necessitates a change in U.S. policy, noting that emerging foreign laws “were generally consistent” with the approach the Office is taking. That said, the global landscape is not entirely uniform (as discussed in the next section), and U.S. policy could face pressure if key trading partners or international IP frameworks start moving in a different direction.
One cannot ignore the looming question: Will there ever be a point where AI is so advanced that it challenges the human-centric paradigm of copyright? The report hints that it’s aware of the trajectory of technology. For example, future AI might allow more user control or might itself be argued to show creative “intent.” The Office commits to monitoring developments and doesn’t rule out revisiting its conclusions if warranted. Some policy thinkers have floated ideas like granting a limited “neighboring right” or some form of recognition for AI-generated works to address situations where huge value is created without human creativity. For now, the Office sees “no case” for a sui generis AI right. But policy debates are ongoing in academia and international forums (like the World Intellectual Property Organization, which has held discussions on AI and IP). These debates include considerations beyond authorship, such as the treatment of using copyrighted material to train AI (which Part 3 of the Office’s report series will tackle later), and issues of attribution (should AI works carry disclosures that they are AI-made?) and authenticity (combatting deepfakes, etc., which Part 1 of the report series addressed). In summary, the Copyright Office’s current policy is to apply existing frameworks to AI with careful nuance, thereby avoiding premature or overbroad changes, but the conversation is far from over, and IP lawyers should stay attuned to the policy discourse as it unfolds.
Alignment with Past Rulings and Global Perspectives
The positions outlined in the Copyright Office’s report largely align with past U.S. legal precedent on authorship, while some other jurisdictions have started to chart their own paths on AI-generated works. Understanding these consistencies and divergences provides context for how revolutionary (or not) the Office’s stance is and guides lawyers who may advise clients across different markets.
Consistency with U.S. Precedent
American copyright law has long held that a work must owe its origin to a human author to enjoy protection. This unwritten rule was famously illustrated by the 2018 case Naruto v. Slater, where a crested macaque (Naruto) snapped a “selfie” photograph. The Ninth Circuit held that since Naruto was an animal, not a human, he could not own a copyright; the photo therefore had no valid author and no copyright. That case underscored that under U.S. law, authorship is inherently tied to human creativity. More recently, in Thaler v. Perlmutter (D.D.C. 2023), a federal court confronted AI-generated art directly. The plaintiff had attempted to register an image autonomously created by an AI system (with no human input beyond initiating the process). The court agreed with the Copyright Office’s refusal to register the work, reaffirming that human authorship is a bedrock requirement of copyright and that a purely AI-created piece is ineligible.
The Office’s January 2025 report explicitly leans on this decision, citing it as judicial confirmation of the Office’s approach. Furthermore, the Office’s own administrative guidance and practices, such as the 2023 policy statement on AI in registrations, already reflected this stance, and the report builds upon them. For example, the Office had updated its Compendium to clarify that works lacking human authorship (like those created by animals or purely by machines) are not registrable, and that guidance is now reinforced and elaborated in the report. In essence, the report does not mark a sharp change in U.S. doctrine but rather a reaffirmation and expansion of existing principles in light of new technology. It strives to show that even cutting-edge AI issues can be addressed with the familiar tools of copyright law, namely the ideas/expression dichotomy (prompts being akin to ideas) and the requirement of original expression by a human mind. By anchoring its conclusions in precedent, the Copyright Office is aligning itself with the continuity of U.S. copyright jurisprudence, signaling that creators and courts should treat AI-assisted works under the same analytical framework they would apply to any other work when assessing originality and authorship.
International and Comparative Perspectives
Globally, there is no single consensus yet on how to handle AI-generated works, but many jurisdictions share the view that human creativity is fundamental to copyright. The report acknowledges that no foreign court of last resort has definitively recognized AI as an author either. The European Union, for instance, requires that a copyrighted work be the author’s own “intellectual creation,” which has been interpreted to imply human intellect; EU authorities have indicated that AI-generated outputs without human creative control would not qualify for copyright under that standard (though the EU has not needed to litigate a clear AI authorship case so far). In line with that, most European countries would likely treat purely AI-made works as unprotected.
The United Kingdom presents an interesting twist: UK law explicitly addresses “computer-generated works” in its Copyright, Designs and Patents Act 1988. Under Section 9(3) of that Act, if a work is generated by a computer in circumstances such that there is no human author, the statute designates the author as “the person by whom the arrangements necessary for the creation of the work are undertaken.” In simpler terms, UK law allows copyright for AI-generated works by assigning authorship to the individual who took the steps to produce it (for example, the user who entered the prompt or configured the AI, arguably). The duration of protection for such works is shorter (50 years from creation, instead of life plus 70 years). The UK government recently reviewed this provision in light of modern generative AI and decided to retain it, essentially affirming that UK law can accommodate AI-generated outputs by finding a human to assign them to. However, even in the UK, there is debate about what exactly constitutes the “arrangements” and who qualifies as the author in an AI context, is it the end-user? the programmer of the AI? There’s some ambiguity, and as of now, it appears that if a human meaningfully engages with the AI tool (through prompts or setting parameters), the UK would likely consider that human the author of the resulting work. This approach diverges from the current U.S. approach, which would say that without human creative control, there is simply no author. The U.S. Copyright Office did not adopt a UK-like rule; instead, it views the absence of a traditional human author as leaving the work unprotected altogether. For U.S. lawyers, this means that an AI-generated piece might be copyrightable in the UK (with the user as the deemed author) even though it’s not protectable in the U.S. This disparity is something multinational companies must be mindful of, content strategy and rights management might differ across borders.
Another jurisdiction at the forefront is China, which has seen a couple of important court decisions on AI works. Chinese law requires a work to be an “original intellectual achievement,” but it doesn’t explicitly demand a human author. In 2019, a court in Shenzhen (Tencent v. Shanghai Yingxun) held that an AI-generated financial news article (produced by Tencent’s “Dreamwriter” AI) was entitled to copyright protection. The rationale was that the output wasn’t entirely the result of the AI’s own devices – human engineers had provided input in the form of selection and arrangement of data and setting conditions, which imbued the process with human intellectual activity. Effectively, the court found that there were human contributions behind the scenes sufficient to consider the work original. The copyright in that case was attributed to the company (Tencent), since the humans involved were part of the company’s creation process. More recently, in 2023, the Beijing Internet Court examined an AI-generated image made using Stable Diffusion. The court ruled that the image met the threshold of originality and was a protectable work because the user had made a creative contribution through the prompt text and parameter settings, which resulted in a personalized form of expression in the output. Notably, the Chinese court explicitly decided that the user (the person who input the prompt) was the copyright owner of the AI-generated image (not the AI developer), reasoning that the user was the one who intended to create that image and determined the inputs, whereas the platform only provided the service. This approach is more generous to AI-generated works than the U.S. approach: it suggests that a well-crafted prompt and user direction can suffice to confer authorship on the user, at least under Chinese law as interpreted by these courts. However, even China hasn’t gone so far as to say an AI itself could be the legal author; they still find a human actor (user or company) to whom the work’s creativity is attributed. These developments show a potential divergence: in some jurisdictions, courts or statutes are stretching existing concepts to find human authorship in AI outputs (like UK’s deemed author, China’s focus on user input), whereas the U.S. remains more strict in requiring a direct human creative role in the generated expression.
Internationally, discussions continue in forums like WIPO about whether there should be harmonized rules or at least guiding principles. Many countries are likely taking cues from each other as they update guidelines. For example, Canada and Australia currently also maintain that human authorship is required (with Canada’s Copyright Act using the word “author” which courts interpret as implying a natural person, and Australia’s law similarly interpreted by its copyright office). Some countries in the Asia-Pacific are studying the Chinese and UK experiences. The Copyright Office’s report explicitly noted that its recommendation against new protection was in line with what it saw globally, implying that adopting a different approach could isolate the U.S. in an untested regime. For IP lawyers advising multinational clients, the key is to recognize these jurisdictional differences. A piece of content generated substantially by AI might be free for anyone to use in the U.S. (no copyright), yet infringing in the UK if someone else tries to copy it there, since UK law might consider it protected and owned by the original user. Conversely, a U.S. client might have no U.S. rights in an AI graphic but could possibly register it in the UK or other countries that allow it, though enforcing those rights, and figuring out authorship for registration purposes, could raise its own complexities. The global trend seems to be cautious about fully embracing AI authorship, but some jurisdictions are experimenting at the margins with accommodating AI-created works. The U.S. stance as of the 2025 report is firm in its orthodoxy: only human creativity generates copyright, which keeps U.S. law aligned with the majority view internationally, albeit not universally all.
Practical Takeaways for IP Lawyers on AI-Generated Works
- Evaluate and Document Human Contributions: When a client brings an AI-assisted creation to you (be it artwork, text, music, or code), begin by breaking down how the work was made. Identify which parts were generated by an AI and what the client (or other humans) contributed. This might involve reviewing prompt texts, any source material supplied to the AI, and any edits or curation done after generation. By documenting these details, you can assess what elements are likely protectable. For any significant AI involvement, advise the client to save records, as this will be invaluable for registration and enforcement. The Copyright Office has stressed that it will look for a human creative spark in the final work; your job is to pinpoint where that spark lies (for example, the client’s original selection of output A over output B, their combining of AI outputs into a larger work, or their post-generation alterations).
- Be Candid in Copyright Applications: When registering a work that has AI-generated components, ensure the application accurately reflects the human authorship. This usually means including a disclaimer or exclusion for the AI-generated material. For instance, in the “Author Created” field you might list “text and editing” or “overall selection and arrangement,” and in the “Limitation of Claim” section, exclude “AI-generated images” or similar. As the Office’s March 2023 guidance highlighted, failing to disclose AI content can jeopardize the registration. An accurate registration will protect what can be protected (the human-authored aspects) and won’t give a false impression that the entire work is copyrighted. Remember that a registration is a sworn document; knowingly misrepresenting authorship (omitting the AI’s role) could not only risk cancellation of the registration but also raise issues of fraud on the Copyright Office. It’s better to secure a narrow but truthful registration than to claim broad rights that can be easily invalidated. Going forward, the Office may refine the forms or the Compendium to expressly ask about AI and staying updated on those procedural changes will help you file correctly.
- Adjust IP Strategies for AI-Heavy Content: If your client’s business relies on content that is primarily AI-generated (for example, a marketing firm using AI to generate social media graphics, or a game developer using AI to create background music), counsel them on the limitations of copyright protection for that content. They should understand that such works might effectively be part of the public domain in copyright terms. This doesn’t mean the content is without value, but it means the company cannot stop a competitor from copying it purely on copyright grounds. To mitigate this, consider alternative protective measures: does the content incorporate any trademarks or logos that are protected? Is the content kept confidential or behind a paywall (where contract law or anti-scraping laws might provide some shield)? Perhaps the firm can use AI outputs as a starting point but then have human designers refine them to create a protectable final product, a workflow that adds a human creative layer. In some cases, clients might opt for quantity and speed of AI content over exclusivity; that’s a business decision, but it should be an informed one. As their lawyer, outline the trade-off clearly: AI can produce rapid content, but exclusive rights come only with human creative input.
- Incorporate AI Clauses in Agreements: The new landscape suggests adding specific clauses in contracts concerning the use of AI in commissioned works or licensed content. If you’re drafting a work-for-hire agreement or a freelance design contract, you may include a representation that the creator will disclose if generative AI was used and ensure that any necessary human contribution is present for the work to be copyrightable. A publisher or producer might want assurances that the delivered work is eligible for copyright protection (i.e., not purely AI-generated) or else they know which parts aren’t protected. On the flip side, if a client is licensing an image or text and they suspect it might have been AI-created, they might negotiate warranties or indemnities in case it turns out to be unprotectable or infringing on someone else’s rights (there’s also the concern that AI outputs could inadvertently plagiarize training data, but that’s a separate infringement issue beyond authorship). The main point is: don’t ignore AI in contractual frameworks. Much like we have open-source software clauses or morality clauses, we may now see “AI usage” clauses to ensure clarity on authorship and rights.
- Keep Abreast of Ongoing Developments: The copyright status of AI-generated content is a fast-moving area. The Copyright Office itself is continuing its study – Part 3 of its report (forthcoming) will deal with issues of training data and infringement liability, which could influence how AI tools are built and used. Legislators are also paying attention; it’s conceivable that Congress could intervene if they feel the need to adjust the balance (though, as of the 2025 report, the recommendation was that no change is needed). Court cases are likely to emerge as early test beds, and beyond Thaler, we might see disputes over partially AI-created works in music or film. Globally, laws might start to diverge more significantly. As an IP lawyer, staying informed through continuing legal education, following Copyright Office announcements, and even international IP news will enable you to give the most up-to-date advice. Clients will look to you not just for what the law is now, but where the trends are heading. For example, if a client is expanding into the UK or China, you should be ready to explain how their AI-generated assets might actually receive protection there (and thus could be valuable IP in those markets) even if they have none in the U.S. This holistic view can guide decisions like where to register or how to exploit content in different regions.
- Emphasize the Human Element: From a more advisory or even ethical standpoint, you might encourage clients to view AI as a tool to empower human creators rather than a replacement. The Copyright Office’s report implicitly champions human creativity, a stance that clients in creative industries should note. If a client’s project is on the borderline of being too automated, consider recommending involving a professional artist or writer to collaborate with the AI, thereby not only strengthening the copyright position but potentially improving the quality and originality of the work. In other words, find the right balance between innovation and legal protection. Many organizations are now developing internal policies on generative AI use; as their counsel, you can help craft guidelines that align with copyright law (for example, a policy might require that all AI-generated content used externally must be reviewed or modified by a human creator before release, ensuring there is a human author to claim). This reduces legal uncertainty and reinforces to creative teams the importance of their own contributions.
Conclusion
The U.S. Copyright Office’s January 2025 report on AI-generated content sends a clear message: copyright law, as it stands, is equipped to handle AI, provided we hold firm to the principle that a human mind must be behind the expressive elements of a work. For intellectual property lawyers, the report is both a guide and a reminder. It guides us on how to counsel clients through the maze of registering and protecting works that involve AI, and it reminds us of copyright’s fundamental purpose of rewarding human creativity.
As AI technology continues to evolve in astonishing ways, the legal landscape will undoubtedly face new questions. Yet, for now, the balance struck by the Copyright Office is to be inclusive of AI as a helpful tool but exclusive of AI as an independent creator in the eyes of the law. By staying informed of these developments, ensuring accurate legal filings, and proactively addressing the unique issues AI brings, IP attorneys can help clients harness AI’s benefits without falling foul of the limitations of copyright protection. In the grander scheme, this ongoing dialogue between law and technology exemplifies how enduring principles can adapt to novel situations, and it will be fascinating to watch (and influence) how that dialogue unfolds on the global stage in the years to come.