For a U.S. business using generative AI, the safest starting point in 2026 is simple: copyright does not automatically attach to material just because a person typed the prompt, paid for the tool, or selected the final output. U.S. copyright law still centers on human authorship. AI can be part of a protected creative process, but protection generally reaches the human-authored expression, not material produced by the system without sufficient human creative control.
That distinction matters for marketing copy, product images, illustrations, reports, software documentation, and other assets businesses increasingly create with AI. Instead, the practical question is: which expressive elements were actually created by a human?
The 2026 rule: human authorship is still required
The U.S. Copyright Office’s current position is that purely AI-generated material is not copyrightable. Its 2025 report on copyrightability said generative AI output may receive protection only when a human author has determined sufficient expressive elements. AI assistance does not disqualify a work, but the human contribution must itself amount to protectable authorship.
In Thaler v. Perlmutter, the U.S. Court of Appeals for the D.C. Circuit held in March 2025 that the Copyright Act requires a human author and affirmed the refusal to register an image identified as autonomously created by an AI system. In March 2026, the U.S. Supreme Court declined to review the case, leaving that appellate ruling in place.
That does not settle every AI content copyright question. Thaler involved a work presented as having no human author. Harder disputes remain over how much human direction, editing, selection, or control is enough when a person uses AI as part of a creative workflow.
Are prompts enough to claim copyright?
Usually, not by themselves under current US Copyright Office AI guidance. The Office concluded that, given how generative systems operate, prompts generally function as instructions that communicate desired ideas or attributes rather than giving the user direct control over the exact expressive details that appear in the output.
This is especially important for AI art ownership. A person may spend hours refining prompts, regenerating images, and selecting one result. That effort can be commercially meaningful, but effort alone is not the legal test. Copyright focuses on authorship of the final expression.
The boundary is still developing. A pending federal dispute involving artist Jason Allen and his Midjourney-created work has focused attention on whether extensive prompting and post-processing can establish sufficient human authorship. Until courts provide clearer rules for those facts, businesses should avoid assuming that sophisticated prompting by itself creates an enforceable copyright in the resulting image.
What parts of an AI-assisted work can be protected?
Protection can exist when human-created elements are sufficiently original. A writer might use AI to generate rough text but then independently rewrite passages, develop the structure, and add original analysis. An illustrator might start with an AI image and substantially redraw characters, backgrounds, lighting, and composition. A designer may also create copyrightable selection, coordination, or arrangement when combining human and AI material in an original way.
The key is that copyright covers the human-authored material. It does not necessarily expand to every AI-generated element included in the finished product.
This is why a business should distinguish between using AI as an assistant and allowing AI to determine the expressive result. The more the final work reflects identifiable human creative choices, the stronger the argument for protecting those human contributions.
A practical business example
Suppose a company asks an image generator for a packaging illustration and publishes the first acceptable output unchanged. Even if an employee wrote a detailed prompt, the company may have little or no copyright protection in the image itself under the Copyright Office’s current approach.
Now change the workflow. A staff designer generates several concepts, combines portions of two drafts, redraws the main subject by hand, creates an original background, adds custom typography, adjusts the composition, and saves layered working files documenting those choices. The AI-originated portions may still need to be excluded from a registration claim, but the designer’s original additions and arrangement may be protectable.
The actionable lesson is to document human authorship. Keep source files, revision history, design layers, drafts, editorial notes, and records showing who made the creative decisions. This will not manufacture copyright where none exists, but it can help identify and support the human-authored material that does qualify.
Registration requires accurate disclosure
The Copyright Office’s registration guidance says applicants should identify the human authors and describe the human-authored contribution. AI-generated material that is more than minimal should generally be excluded from the claim. Applicants should not list the AI system or its provider as an author merely because the technology was used.
For a mixed work, that may mean claiming human-authored text, visual edits, or selection and arrangement while disclaiming AI-generated images or passages. Businesses preparing valuable assets for registration should consider having an intellectual property attorney review the authorship record and application language.
Tool terms do not replace copyright law
Another common source of confusion is the difference between contractual rights and federal copyright. An AI platform’s terms may say that, between the provider and user, the user receives certain rights in outputs. But a contract cannot make otherwise uncopyrightable machine-generated expression qualify for federal copyright protection.
For businesses, generative AI legal risk has several layers: copyrightability, infringement, licensing terms, confidentiality, vendor restrictions, and ownership rules in employee or contractor agreements. Treating “the platform says I own it” as the entire analysis is risky.
FAQ
Can I copyright an image made with AI?
You may be able to protect original human-authored elements of an AI-assisted image, but a purely AI-generated image is not protected under the Copyright Office’s current position. The result depends on what expressive elements a human actually created.
Does writing a detailed prompt make me the copyright owner?
Not necessarily. The Copyright Office has said prompts alone generally do not provide enough human control over the resulting expression. More direct human creation, modification, or arrangement can change the analysis.
Can a business register a work that contains AI-generated material?
Yes. Mixed works can be registered when they contain copyrightable human authorship. The application should identify the human contribution and appropriately exclude AI-generated material from the claim.
Is AI-generated content automatically public domain?
That phrase can be misleading. Material lacking copyright protection may be free from a copyright claim in that material, but other rules can still matter, including contracts, trademarks, privacy, publicity rights, and laws governing particular content or uses.
What businesses should do now
AI-generated content copyright rules are becoming clearer at the edges but remain unsettled in the middle. The firm rule is human authorship: a machine cannot be the sole copyright author under current U.S. law. The open questions concern how much human control and creative transformation are enough in AI-assisted works.
Businesses can reduce uncertainty by using AI as a tool rather than a substitute for authorship, preserving evidence of human creative work, reviewing platform terms, and making accurate disclosures when registering mixed works. For high-value brand assets, campaigns, software, or artwork, a documented human-led workflow is far safer than relying on prompt history alone.