In the last installment of this series, we looked at what it actually means when a law firm says it uses AI — and why the answer varies enormously from firm to firm. A number of readers came back with a version of the same follow-up question, but from a different direction. Not: how are my lawyers using AI? But: my company has been using AI to generate product designs, marketing materials, and a new brand identity. I assumed we owned all of it. Do we?
It is a reasonable assumption. It is also, in many cases, wrong.
Imagine your marketing team uses an AI tool to generate a library of product images, advertising copy, and campaign visuals for a major product launch. The team spends weeks refining prompts, selecting the best outputs, and building the campaign around them. It is polished, original-looking work. You spend real money rolling it out. Six months later, a competitor lifts the images and copy wholesale and runs their own version. You call your lawyer. Your lawyer calls back with news you did not expect: because the work was generated by AI, you may not have the copyright protection needed to stop them.
That scenario is playing out in businesses right now. The intellectual property laws that protect creative work, inventions, and brands were written for humans. AI has arrived faster than the law has adapted, and the gap between what businesses assume they own and what they can actually protect is significant.
Here is where copyright, trademark, and patent law each stand.
Copyright: The Human Authorship Problem
Copyright protects original creative works — photographs, illustrations, written content, music, software, and similar material. Before getting into what AI does to copyright, it is worth noting what copyright does not protect even when a human being creates it: short phrases, titles, names, slogans, and most logos. These are generally not copyrightable regardless of how creative or distinctive they are. That is not an AI problem — it is simply the boundary of copyright law. Protection for logos and brand identifiers comes from trademark law, which we will get to in a moment.
Within the territory copyright does cover, AI has created a significant problem. The U.S. Copyright Office has been clear: it will not register works generated entirely by AI. The D.C. Circuit affirmed that position in 2025, holding in Thaler v. Perlmutter that AI systems cannot be “authors” under the Copyright Act. The court noted that the statute’s language — its references to widows, children, heirs, and bequests — reflects a foundational assumption that authors are people. An AI tool, however sophisticated, is not.
That means AI-drafted marketing copy, AI-generated product illustrations, AI-composed music, AI-written training materials — none of it is automatically protected just because you created it with an AI tool and paid for the subscription.
But that is not quite the end of the story. The Copyright Office has also said that genuine human creative choices made in the process can be protected. If you evaluated hundreds of AI-generated options and selected, arranged, and modified elements in a way that reflects real creative judgment, those human contributions may be protectable — even if the underlying material came from a machine. The copyright covers what you did, not what the AI did.
What this means practically: an image taken as-is from a single prompt gives you nothing to register. A campaign assembled through weeks of selection, refinement, and creative decision-making may give you something. The more demonstrable human authorship in the process, the stronger the potential claim. The less, the weaker — down to zero.
The business consequence is real. If you cannot copyright your AI-generated content, a competitor can copy it freely. There is no infringement where there is no protected work. Companies have already encountered this, and some have lost the ability to stop copying they assumed they could prevent.
Trademark: Better News, With One Caveat
Trademark law works differently, and here the news is better. Trademark protects brand identifiers — names, logos, slogans — that distinguish your goods or services from someone else’s. Crucially, trademark law does not require individual human authorship. What it cares about is not how an identifier was created, but whether you are using it in commerce and whether it functions as a recognizable brand.
That means an AI-generated logo can generally be trademarked. The U.S. Patent and Trademark Office does not ask whether a human or a machine designed the image. It asks whether the image serves a branding function and whether it is sufficiently distinctive to qualify for protection. If you have been using an AI-generated logo consistently in your business, you may already have common law trademark rights in it — and federal registration would significantly strengthen your ability to enforce them.
This is the most accessible form of protection available for AI-generated creative work, and it is underused. Businesses that have built a brand around AI-generated visuals should seriously consider registration, even if copyright is unavailable.
The caveat is this: trademark and copyright protect different things, and losing one weakens your overall position. Trademark protects the branding function — it stops competitors from confusing your customers. Copyright protects the creative work itself — it stops anyone from reproducing the image regardless of context. Without copyright backup, someone who copies your AI-generated logo for use in a completely different industry may be harder to reach. The two forms of protection are designed to work together, and the absence of one leaves a gap.
One procedural note: the USPTO has signaled that applicants should be transparent about the role AI played in creating the work being registered. Omitting or misrepresenting that information can affect the validity of your registration down the road. And while they do not currently ask whether a human or machine designed the image, the USPTO has signaled that may change
Patent: The Inventor Problem
Patent law presents the sharpest challenge of the three, and the one most likely to affect businesses that use AI for product development or technical innovation.
The Patent Act requires that every patent identify the actual human beings who invented the claimed technology. Not the company. Not the software. The people. The Federal Circuit made this explicit in Thaler v. Vidal, rejecting an attempt to list an AI system as the named inventor of two novel devices. The court held that the statute’s reference to “individuals” means natural persons, and that AI systems do not qualify.
The practical question for most businesses is not whether an AI can be a named inventor — it cannot — but whether the humans working with AI can still qualify as inventors when the AI is doing significant analytical or generative work. The USPTO’s current answer is yes, provided the human made a “significant contribution” to the actual conception of the invention. Using AI to run experiments, evaluate options, or model outcomes does not disqualify a human inventor, as long as the human genuinely contributed to what the invention is, not just how it was built.
The harder question is where that line falls. If an engineer prompts an AI system and the AI generates a novel solution the engineer then implements, who conceived the invention? If the core idea came from the machine and the human’s role was primarily execution, the claim to inventorship — and the validity of any patent — becomes vulnerable.The relevant question is who first conceived the idea — who formed in their mind the definite and permanent idea of the complete and operative invention. A person who frames the conceptual problem the AI is solving, who specifies the use case, or who originates the inventive concept the AI then drafts is the inventor for purposes of the statute, even where the AI generated the language used in the application. A person whose role is limited to selecting among AI-generated options without contributing to the underlying concept is not. Courts and the USPTO are still working through the details, and the law will continue to develop as more AI-assisted patent applications are tested.
What is already clear is that documentation matters enormously. Companies developing products with AI assistance should be keeping records of who did what, at what stage, and what judgments the human contributors actually made. That record is what separates a defensible patent from one that can be challenged on inventorship grounds. Where inventorship is wrong on a pending or issued patent, federal regulations permit correction without abandoning priority — but correction generally requires the cooperation of the named inventor, and inventorship disputes that reach litigation can result in invalidation rather than correction.
What to Do Right Now
The law in all three areas is still catching up to the technology. But the gap between what people assume they own and what they can actually protect is not theoretical — it has consequences in licensing, acquisition, investment, and litigation. A few steps reduce your exposure and preserve your options.
Document your process. If you are creating logos, product designs, written content, or marketing materials with AI, keep records of your prompts, the options you evaluated, and the choices you made. The difference between protectable and unprotectable human authorship may come down to what you can demonstrate about your creative involvement.
Register your trademarks. Trademark registration is the most straightforward protection available for AI-generated branding, and it is surprisingly underutilized by businesses that have built their visual identity with AI tools. If you are relying on an AI-generated name or logo to identify your business, registration gives you the tools to enforce it.
Be careful about representations of ownership. Telling a licensee, an investor, or a buyer that you “own” AI-generated content may be accurate, or it may be overstated, depending on how the work was created. That distinction matters when the representations appear in contracts or due diligence disclosures.
Get clarity before you have a problem. If your business model, your product, or your brand depends on protecting something created with AI, the time to understand what you actually own is before a competitor copies it — not after.
The Gap That Needs Closing
AI tools are sold as creative partners. The interface feels like collaboration. The output feels like yours. The law does not always agree, and the gap between that feeling of ownership and the legal reality is where disputes are forming.
None of this means you cannot build a protectable business around AI-generated work. Many companies are doing exactly that. But it requires understanding which protections are actually available, which assumptions are wrong, and where the law is still unsettled. The companies that figure that out early will be better positioned than the ones who discover the limits of their ownership at the worst possible moment.
As we noted in the last post, the lawyers best positioned to help with questions like these are the ones who understand AI not as a novelty but as a working part of their practice. The IP implications of AI-generated work are exactly the kind of issue that gets missed when legal advice on AI comes from someone who has not thought seriously about how these tools actually work.
If you have questions about protecting AI-generated work product in your business — or about what your current contracts and registrations actually cover — we are glad to have that conversation.
A Word About SilverCain
Silver Cain PLC was founded on the premise that businesses deserve both exceptional litigation experience and direct partner access — and that you should not have to choose between them. Leon Silver and Rebecca Cain have spent decades handling the most complex business and real estate disputes in Arizona and nationally. If you are evaluating counsel in Phoenix, we welcome the conversation.
