Generative AI forces rethink on IP law

As artificial intelligence tools become more powerful and more widely used, Missouri attorneys will face increasing questions about what that means for intellectual property.

Clients are seeking guidance on how to protect proprietary information when using tools like ChatGPT and whether AI-generated innovations are eligible for protection.

Though the U.S. Copyright Office and the U.S. Patent and Trademark Office have issued early guidance, attorneys say much of the legal doctrine remains unsettled. With new lawsuits and legislative proposals emerging, Missouri lawyers are preparing for changes that could reshape IP law.

Patent protection limits

As artificial intelligence becomes more sophisticated, patent attorneys are confronting new challenges over whether AI-assisted inventions can, or should, qualify for patent protection.

The issue has surfaced in cases involving AI systems like DABUS and Google.

In 2022, the U.S. Court of Appeals for the Federal Circuit determined in the case of Thaler v. Vidal that under the Patent Act, an “inventor” must be a natural person, thereby excluding AI systems like DABUS from being recognized as inventors. The ruling stemmed from a challenge by Stephen Thaler, who listed DABUS as the sole inventor on two patent applications.

“They (the courts) have all reached the same conclusion that merely using artificial intelligence to invent something is not patentable,” said Rudy Telscher of Husch Blackwell. “A real person has to be involved to some extent.”

The question is playing out in courts across the country. In April, the Federal Circuit affirmed a Delaware court’s invalidation of machine learning patents held by Recentive Analytics, ruling the claims were abstract and ineligible.

The question of patent infringement by AI is also being addressed in the judicial system. Google’s $1.6 billion settlement with Singular Computing in January over AI chip patents highlights how infringement issues are gaining traction.

The USPTO released its “Artificial Intelligence Strategy” in January, outlining IP policy goals and investments in infrastructure to support responsible AI use.

“I don’t see the patent office ever saying, you know, this patent is ChatGPT,” said Stinson IP attorney Steven Cosentino. “This is one that hasn’t fully played out yet, but where it’s getting a lot of focus is … does the fact that a human had to enter something into a prompt to trigger the AI tool … is that sufficiently created enough?”

Telscher said his advice to clients using AI is “as best as you can, you’re documenting what AI did.”

“The courts may say we’re not going to give protection for that part what was AI-driven,” Telscher said. “Just because you can get a patent at the front end that had AI involved, what happens when you go to try to enforce it, where you got to show what AI did and what you did … Companies that are using AI as part of the process … (should) make sure they’re doing their best to document what the inventor’s contributions were.”

Copyright’s gray areas

Questions about copyright have become some of the tougher legal issues raised by artificial intelligence, with courts determining who owns the work created with the help of AI, and how infringement applies when AI is involved.

The U.S. Copyright Office launched its AI initiative in early 2023, and maintains that under current law, copyright protection requires human authorship. Purely AI-generated works aren’t copyrightable, and detailed prompts alone don’t qualify. In works containing both human and AI-generated content, only the human-created portions are eligible for protection. If AI is used as a tool to support the creative process, the resulting work may still qualify for protection.

Proving direct infringement remains a hurdle. Plaintiffs must show the AI output is substantially like the original, a complex task given how generative AI models function.

“It’s usually very difficult to be able to specifically tie the output back to any one specific input, because the relationship between the two is just so attenuated by the data model that sits in between (the input and output), and by the user input that’s used to generate that specific output,” said attorney Benjamin J. Siders of Lewis Rice. “If I go to it and say, give me an image of somebody playing an accordion at an Oktoberfest festival … there’s a million pictures of that out there. How do you know which specific one, which accordion, what was used … Nobody owns a copyright to the idea of an accordion and (Oktoberfest). It’s only your particular expression.”

As lawsuits continue to work their way through the courts, Siders said the outcome may hinge on whether courts find AI companies or their users ultimately responsible.

“That’s going to be the challenge … proving that the copyright work was used to train the engine at all, if it was, that it’s not a fair use, and then if it’s not, that what was produced as an output is infringing, and if it is, who’s responsible for that,” he said. “The image was generated by the AI company, but they didn’t provide the prompt or the input, just the underlying data that was used to assemble it … It’s like … I’m selling a book full of construction paper, and somebody cuts it all up and reassembles it into … a collage that looks like Mickey Mouse, well, who’s the infringer there?”

Branding and AI risk

Beyond content ownership, attorneys say artificial intelligence is also creating new risks for brand protection, as AI-generated marketing and media raise concerns about trademark misuse and consumer confusion — and who’s to blame.

One notable case involves Tesla, Warner Bros. Discovery and Alcon Entertainment, the studio behind Blade Runner 2049. Alcon alleged the unauthorized use of AI-generated imagery resembling some of the movie’s visuals during a promotion for Tesla. However, a federal judge tentatively dismissed the trademark violation claims, noting there was no direct competition between the parties. However, the court allowed Alcon’s copyright claim to proceed.

Attorneys say one of the most common uses of AI in branding is through tools like ChatGPT to generate names, slogans or even logos for business — and that’s where potential conflicts begin.

“One thing I’ve seen most recently are potential trademark applicants using it — like ChatGPT and other programs — to generate ideas for brand names or for taglines,” trademark attorney Morris Turek of St. Louis said. “Then they come to me and say, ‘AI generated this name for this product,’ and we perform a search on that name to see whether its available or not for the product or service that the potential client wants to provide.”

But while these tools can be helpful brainstorming aids, they carry legal risks. Turek noted that AI doesn’t evaluate potential conflicts or legal bars to registration, like whether a name is deceptively descriptive or reserved under federal law.

“That’s actually happening all the time,” he said. “The problem with AI is that it doesn’t know whether there are actually any conflicts with these names that it’s generating … it might generate something like ‘Olympic’ in the name, which only the U.S. Olympic Committee is allowed to register. Or it might create a name like ‘Shooting Star Organic Apples’ when the product isn’t organic — and that would be deceptive trademark.”

AI-generated branding content also risks overlapping with existing trademarks. Attorneys also see risks in how clients use generative AI to prepare trademark applications themselves — often without understanding the legal consequences.

“I also see a potential problem with AI being used to help prepare and complete trademark applications,” Turek said. “Because AI doesn’t know what’s in the applicant’s head … The applicant has to have a legitimate intent to actually use the trademark in connection with those … products.”

Clients who file AI-generated product lists without intent to use them also risk legal issues. Even if approved, AI-assisted registrations may not hold up.

“I see that there could end up being a lot of … applications with information and data that is not accurate. And that could lead to the applications being rejected down the road or challenged by a third party down the road,” he said. “Maybe the application will get through … But is the registration subject to challenge? Is it subject to cancellation for whatever reasons? And the applicant won’t know that until it’s challenged.”

Trade secrets and data privacy

For businesses relying on confidential data and proprietary information, the growing use of generative AI inside the workplace also raises concerns about trade secret protection. Unlike patents, which are eventually publicly disclosed, trade secrets remain confidential under “lock and key,” Cosentino said.

“AI … from a privacy, data security standpoint … anything you put in the engine, you’re potentially losing privacy of confidentiality of it,” he said. “Let’s say I’m a principal at the Coca-Cola company … I want to make something called ‘New New Coke’ … So, I take that Coca Cola formula, and I put it in ChatGPT and say ‘Hey … why don’t you create … New New Coke.” … Well, the moment I put that Coca Cola formula into ChatGPT, I just lost my trade secret because there’s no confidentiality protection.”

General data privacy is also a concern when using open AI systems, said Paule, Camazine & Blumenthal attorney Grant M. Gamm.

“You’re giving that information to Google, to other huge companies which have a track record of not being the most ethical users of privileged information,” Gamm said.

The European Union has already taken steps toward addressing those concerns with the AI Act. Among its key provisions are transparency requirements for generative AI models, including when content is AI-generated, and public summaries of any copyrighted materials used in training. Attorneys say the act may offer a preview of how regulators in the U.S. could begin demanding more transparency and accountability around how AI systems are trained.

Policy and practical use

Despite the focus on risk, attorneys say AI is already helping clients with IP enforcement.

“One of the cool things about AI is that it’s really good at going out and scouring different sources … and finding whether there are similarities,” Cosentino said. “There’s some real positives from an IP enforcement perspective. Same thing with articles. You can tell AI to go see if there’s a plagiarized article out there.”

Polsinelli also launched its in-house AI-tool, PatentCAD, in 2021 to help its attorneys draft patent applications more efficiently and consistently. The firm said the software improves quality and saves time, particularly for new lawyers.

At the same time, companies are increasingly seeking legal guidance on how to use AI responsibly.

“We’re now starting to see clients asking for AI policies, how they should be using AI, what considerations should be taken into account,” Gamm said.

Still, new legal protections under consideration could reshape how IP is enforced. The proposed NO FAKES Act would create a new category of intellectual property law covering a person’s digital likeness, making it unlawful under certain circumstances to create photorealistic AI-generated images of real people without their consent.

“It’s mainly to stop people from making deepfakes of celebrities that misrepresent them or lie about them,” Siders said. “It can also be used to minimize political manipulation, making fake speeches from political figures.”

Past precedent, new tech

Though artificial intelligence may feel like uncharted territory, legal experts say the courts are not starting entirely from scratch. When it comes to copyright protection and authorship, some foundational questions have already come up, like the infamous 2018 case where PETA sought copyright protection on behalf of a monkey who snapped a photo. The court said no.

“A lot of the issues that (AI) brings up are not different in category from prior technical challenges to our IP regimes,” Siders said. “A lot of the answers we’re getting now are things that copyright lawyers kind of expected … because it’s how the law already works.”

While AI capabilities are advancing quickly, attorneys caution that it hasn’t yet reached the level of general or superintelligence, and legal systems still have time to adapt.

“Right now, we have artificial intelligence, that’s where we’re at,” Telscher said. “We’re not at AGI yet. Humans are still more creative. But AI is getting better and better with every passing month.”

AI has long existed in everyday tools like predictive text and spellcheck, but ChatGPT and other generative systems pushed it into public view, accelerating how companies adopt automation.

“We talk about AI like it’s this new thing, and it’s really been around a lot in different forms,” Cosentino said. “But ChatGPT makes it accessible to a much larger group of people. Now companies are trying to figure out how to respond.”

Attorneys say while the legal system may lag behind AI’s rapid development, businesses, policymakers and courts are watching closely.

“All we know is it’s getting better and better with every passing month, and certainly every year,” Teschler said.