How Can Businesses File Copyright Claims for AI-Generated Product Images?

What Is the Copyright Status of AI-Generated Images?

As of September 29, 2026, a business generally cannot obtain U.S. copyright protection for an image created solely by prompting an AI image generator. The U.S. Copyright Office requires human authorship, meaning protection generally depends on whether a person controls the expressive elements of the work rather than merely describing an intended result. The distinction is not simply whether a business owns the computer, account, subscription, or final image file. What matters is how much original human expression is reflected in the final work, although the Copyright Office has not adopted a precise percentage threshold for determining when AI-assisted output becomes protectable.

Also worth reading: What is the AI image compliance checklist and how do businesses ensure generated visuals meet legal and platform standards in 2026? · Do You Immediately Own Copyright to Photos, Videos, and AI-Generated Content? · What are the ai generated video copyright rules in 2026?

A text prompt such as “a premium red coffee bottle on a marble table” ordinarily specifies subject matter and style but may not, by itself, constitute authorship. The result becomes more defensible when a human artist manually paints, draws, photographs, or substantially redesigns identifiable elements. A business can also create a protectable compilation if it selects, coordinates, and arranges individually unprotectable images into an original product catalog, advertisement, or website display, but that protects the selection and arrangement rather than automatically giving ownership of every underlying AI-generated image.

A useful legal warning sign is that you cannot clearly separate what the AI produced from what your team created. If the final image was generated in a few prompts without meaningful human revision, describing it as copyrightable is risky. Prompting can contribute authorship in some circumstances, but the current Copyright Office guidance emphasizes the nature and degree of human creative contribution, not ownership of the tool or payment of a subscription.

The controlling question is whether a human authored the expressive work, not whether AI technology merely assisted that authorship. A business can use AI to propose alternative compositions, yet the person directing and refining the image may still be the author if the final result reflects sufficiently controlled human choices. Conversely, high prompt detail does not automatically settle the issue, because a model may make unpredictable additions, replace requested details, and introduce visual elements the human did not create.

The recurring “human authorship” dispute involving artist Jason Allen illustrates why courts and agencies focus on the creative process. Thaler argued that an AI system should be considered the author of a work entered through the U.S. copyright system. The Federal Circuit rejected that position, and the Supreme Court declined to hear the case in March 2023. In 2025, the same judicial approach was applied in the Allen dispute, and the Supreme Court was reported to have denied further review in March 2026. Those decisions reinforce the existing principle that a machine cannot qualify as a human author, although they do not create a complete rule for every hybrid creative process.

For businesses, the distinction between assistance and autonomous creation is factual. If a designer creates a custom product photograph with AI, masks out the original, adds a logo, colors the background, and composes the product, the strongest protection may cover the edited or composite version. The business should preserve the original photograph, editing files, masks, layers, and export history because they demonstrate human control. It should also document prompts, rejected generations, brushwork, transformations, and design decisions. A clean final image by itself provides little evidence of which features qualify as human-authored.

There is no universally accepted “more than 50 percent human-made” rule, and the Copyright Office has not announced a bright-line threshold. Statements that 10%, 50%, or 90% AI involvement automatically determines copyrightability should therefore be treated with suspicion. The legally important inquiry is more demanding: the human must contribute original expression that can be evaluated apart from the AI-generated material. The U.S. Copyright Office’s 2025 report on copyrightability involving AI also cautioned against simplistic metrics and emphasized that human contribution is assessed work by work.

How to Create Images That Have a Better Chance of Protection

The safest approach is to begin with a photograph or manually created illustration. AI can remove an imperfection, expand a background, suggest a layout, or create an internal concept, after which a human designer makes controlled edits. The stronger the human contribution, the easier it usually is to identify and separate the author’s work. Merely rerunning a model with a slightly different prompt, applying a broad automatic upscaler, or resizing an image usually adds much less to authorship than drawing, painting, arranging, or manually retouching specific elements.

A practical workflow begins by retaining human-created source material. The team should save original product photographs, sketches, typography projects, masks, and layered files separately from AI outputs. Next, it should use AI only for clearly identified transformations, such as generating several background concepts behind a product cutout. A designer should then make visible manual modifications, including compositing, lighting corrections, retouching, text, and selection of specific generated details. The process must be documented with dated project files, design notes, and version histories.

Before registration, the applicant should compare the initial generation with the final image and identify exactly which features were added or changed by a person. Product shape, familiar brand elements, and generic background forms may not be protectable if they are dictated by functionality or generated without meaningful control. Original graphic elements, such as a distinctive illustration, carefully designed arrangement, or modified lighting sequence, may receive stronger protection. This separation should be reflected honestly in the registration application, because incomplete disclosure can create legal problems beyond a later dispute with the Copyright Office.

Companies using AI-generated product images commercially should avoid promising customers exclusive copyright ownership of the entire image. A contract can assign rights in a commissioned work, but a contract cannot manufacture copyright where the law provides none. The agreement should instead explain which human-created materials are being licensed, whether generated outputs are being reused, whether the provider is allowed to submit them for other projects, and whether the customer may modify them. Clear contractual records remain valuable even when the image itself is not registrable.

Filing a U.S. Copyright Registration or Claim

First, determine whether the asset contains enough human-authored expression to register. If the company can show that a designer manually created or substantially revised the image, it may register the human-created work or the protectable composite. If only the prompt and the model created the result, filing a copyright application is unlikely to produce protection. Registration does not create copyright; it is a federal prerequisite for filing most copyright infringement lawsuits in the United States and can affect the remedies available after infringement is proved.

For an eligible work, the applicant should use the U.S. Copyright Office’s online registration system or a paper application. One regular application is intended for a single author claiming a single work that is not a work made for hire; its current federal base fee is $45, though surcharges and additional-claim fees can increase the total. A work made for hire, a group of unpublished or published works, or other filing categories can require a different application and fee. Processing times and portal procedures may change, so the company should verify fees and filing options directly with the Copyright Office.

The application must accurately identify the author or claimant, the title, publication status, and completion year. The claimant should not list the AI provider as the author, and a statement that the work was “created by AI” may be incomplete if substantial human editing occurred. If an AI-generated portion cannot be excluded, the applicant should seek advice about how to describe the work and claim only the necessary human contributions rather than representing the entire output as original. The Office may request source materials or additional information when human authorship is unclear.

A copyright registration should not be confused with a takedown complaint submitted to an image host, marketplace, social platform, or search engine. A registration application goes to the Copyright Office; an infringement claim goes to the party accused of unauthorized use. Platforms have their own reporting processes, deadlines, and evidence requirements, so a company may need both a registration for stronger litigation standing and a separate abuse-reporting process for immediate removal. Removing an image does not resolve ownership, and receiving a platform refund or reinstatement does not decide whether copyright exists.

Responding to Unauthorized AI Reuse of Product Images

A copyright claim is appropriate when the accused copy uses protected human-authored photography, illustration, editing, or compilation in violation of a valid exclusive right. Merely using the same general idea—for example, “a blue bottle on a white background”—is not ordinarily copyright infringement. The stronger concern is often reproduction of a distinctive photograph, a close recreation of an original composite, substantial copying of a catalog’s creative arrangement, or distribution of a modified derivative image.

Evidence should be assembled before contacting the other party. The business should preserve the original file, layer files, timestamps, registration certificate, registration record, source URLs, screenshots, purchase receipts, generator terms, prompt logs, and a side-by-side comparison. The analysis should distinguish exact duplication from resemblance caused by the product’s shape, brand appearance, packaging, or ordinary commercial use. A person familiar with copyright law should also check whether the use qualifies as fair use, a license, or another permitted use.

If there is no protectable image copyright, alternative legal grounds may deserve review. Trademark law can address confusing uses of a brand or logo, but a logo is not automatically copyrightable and trademark protection depends on source confusion or dilution. Right-of-publicity laws may protect commercial uses of a person’s name, image, or voice, while false-endorsement and false-advertising rules can address a fabricated approval. A contract may prohibit automated training, reuse, or vendor competition, and the vendor’s terms may impose ownership or confidentiality conditions even when the image itself is unprotectable.

For AI product images, a hybrid enforcement package is often more sensible than treating one legal theory as universal. The company can register the original human photography, design a copyright claim for that work, use trademark claims where brand confusion exists, and rely on contracts for AI vendors and freelancers. Against a deepfake or synthetic endorsement featuring a real person, right-of-publicity, false-endorsement, and deceptive-practices theories may be more relevant than an image-copyright claim based on an AI-only work.

Costs, Deadlines, and When to Act

The most important deadline for a U.S. copyright owner is the requirement to register before filing an infringement lawsuit. There is no general federal rule requiring registration within a fixed number of days after creation or publication, although prompt registration can preserve proof of title and improve the practical value of a claim. Registration can be filed before infringement begins, after infringement is discovered, or before a future suit in a broader commercial dispute. Early filing is not required for every negotiation, but waiting can complicate evidence and strategic planning.

The federal filing fee for the most common single-author, single-work application is $45, while grouped applications, corrections, special handling, litigation-related submissions, and expedited processing can cost more. An attorney may charge a separate fixed fee, hourly rate, or contingency arrangement; there is no reliable national “AI copyright claim” price because complexity varies. Platform complaints may be free, but they do not replace legal review. Businesses commonly spend more on preserving source files and having counsel analyze human contributions than on the government filing fee itself.

The supplied legal context is set for September 29, 2026, but litigation, agency policy, regulations, and platform procedures can change. A company should verify the current fee and current Copyright Office guidance on the filing date, and it should reassess any claim involving recent precedent or the still-developing law on AI training and output. Trademark and publicity rights also vary by state and country, so a U.S. registration should not be treated as a global remedy. If the asset is sold, licensed, franchised, or advertised internationally, counsel may need to assess each relevant jurisdiction before enforcement.

A business should act before public launch when it creates a campaign that depends on exclusivity, when a competitor or platform begins using a close copy, or when a vendor claims rights in the underlying process. A formal application is more urgent where a future lawsuit is possible because U.S. registration normally must precede the filing of the complaint. For a minor internal concept, preserving records and correcting weak authorship may be more rational than immediately paying a filing fee. For a high-value image, the company should preserve the creation record, register eligible human work, review contracts, and determine whether trademark or publicity claims add value.

Common Mistakes and Better Alternatives

A common mistake is assuming that paying for an AI tool transfers copyright in the generated output. Subscription terms may grant permission to use the image, but that contractual permission is different from exclusive copyright rights. Another mistake is treating the product shown in the picture as owner-authored art; ordinary bottles, shoes, furniture, and packaging may be depicted in ways dictated by the product’s design and function. It is also incorrect to assume that a trademark symbol makes the entire commercial image copyrightable.

Businesses also make the error of registering an image while leaving out generative-AI material. This is not a useful workaround because applicants have a duty to provide accurate information, and the Office may compare the work with submitted records. Other errors include using only automatic generative fill, adding a logo that is smaller than the copied work, and claiming ownership of a prompt without identifying original human expression. Buying an image from a marketplace can help document a transfer of contractual rights, but it cannot guarantee that the seller possessed copyright to grant.

The better alternative is a layered asset strategy. Use real product photography as the core image, manually art-direct the composition, and limit AI to tasks such as background expansion or optional concept creation. Store a human-edited master, a layer-based working file, dated photographs, design notes, and the exact generation inputs. Register the protectable human work when its commercial value justifies the process. Where the image recreates a famous artistic style, a living artist’s identity, a private creator’s photograph, or a recognizable person without permission, permission from that rights holder may be more important than after-the-fact copyright registration.

A Defensible Business Workflow

For a company considering AI product images, the defensible process starts with an ownership audit. Identify the image generator, any stock or commissioned material, human editors, and the party operating the account. Next, make the human contribution visible through original photography, manual retouching, layout, and graphic elements. Preserve creation evidence for the entire project rather than only the exported JPEG or PNG, because a flattened file removes the clearest evidence of separate human and machine contributions.

The company should then have counsel or an experienced reviewer compare the source materials with the final work before registration. If the result is AI-only, the business should not describe it as exclusively copyrighted, but it can still use it subject to the provider’s license and relevant third-party rights. If the work contains human authorship, the application should claim the necessary elements accurately, and the company should not overstate what it owns. If someone uses the image, preserve evidence and submit any platform complaint promptly, but select the remedy based on the legal right involved rather than assuming every unauthorized use violates copyright.

This approach also improves customer confidence without turning every marketing image into a legal project. A product page can state that the company produced the underlying photography and design, licensed any generated component, and reserves rights in its original creative materials. Such wording is not a substitute for ownership, but it is more accurate than claiming full copyright in an autonomous image. The strongest protection comes not from a special AI filing route—there is no separate “AI copyright” form—but from creating and documenting enough original human expression for the established copyright system.