Innovation, AI, and the Law: Manuela Astrid Weixlbaumer on Protecting Intellectual Property

You created an AI-generated image and someone just took it. Do you have any rights?

Let’s say you generate a stunning piece of digital art using AI. It’s unique, eye-catching, maybe even viral-worthy. But then, someone else grabs it, reposts it, or even sells it without your permission. Can you do anything about it? Do you actually own it?

This isn’t just a hypothetical scenario. Can you copyright something made by AI? What happens when AI itself infringes on someone else’s work? And if AI can’t legally ‘own’ anything, does that mean its creations are, in any case, free for the taking?

To break it all down, we sat down with Manuela Astrid Weixlbaumer, a legal professional and IP specialist at a renowned tech company in Austria. Her focus is on trademark law, but she is also fascinated by patent law and the protection of innovation in technical environments.

As an educator in adult learning, she specializes in teaching topics like AI and copyright. Manuela is a member of Women in AI Austria and runs her own YouTube vlog in German, The Cherry Compiler. Her mission is to explain IP-related issues in an understandable way while also giving female inventors the visibility they deserve. Besides, she is also a speaker at panels and conferences.

In this interview, we dig into the biggest opportunities and risks AI presents for patents, trademarks, and copyrights—plus, what every creator, company, and innovator needs to know to stay ahead of the legal curve.

Q. Manuela, thank you for your time. To start, can you give us an overview of how AI is being used today in the areas of trademark, patent, and copyright law?

A. AI is playing an increasingly important role in the field of Intellectual Property. In trademark law, it serves as an upgrade for research purposes. The WIPO (World Intellectual Property Organization) has offered a tool for trademark and design search based on AI for some time now. In patent law, AI makes patent research much easier. It simplifies checking the state of the art and helps determine whether an invention is truly new. After all, that is the key criterion for a patent: It must be new, inventive, and stand out from the existing state of the art.

However, drafting complex patent claims is not something AI can do at this point – it remains a task for humans. We must remember that we are still in the era of weak AI, with Large Language Models. A form of AI powerful enough to independently create patentable inventions would fall into the category of ASI (artificial superintelligence), which is currently still science fiction. If we ever reach that point, existing regulations like the AI Act will definitely no longer be sufficient. And that is, to say the least, an understatement.

In copyright law, the use of generative AI raises many questions rather than offering relief – especially since AI models are trained on publicly available data, which may include content, protected by copyright. This makes it crucial for companies to raise awareness of the potential risks and take steps to manage them.

Q. The European Union has taken a big step in regulating AI with the AI Act, which has been a major topic of discussion from day one, shaping conversations across industries. How do you see this framework influencing Intellectual Property practices, and what challenges might it bring?

A. The AI Act is the „go-to-topic“ in Europe at the moment, as its implementation is ongoing. It aims to provide a legal framework for AI across the entire European Union – such a regulation is unique so far. The goal is to ensure product safety and build trust in AI. The AI Act classifies AI systems by risk categories and sets different requirements depending on the level of risk. Article 4 of the AI Act, which became binding in February 2025, for example, requires companies to train employees on AI literacy. There is no explicit training obligation mentioned, but it’s clear that companies must ensure their employees have the necessary awareness if they are AI providers or deployers. That covers a lot of people.

In the field of Intellectual Property, this means that companies must establish clear processes to ensure AI-generated content is checked and its legal implications evaluated. It’s also worth asking: Do we always need AI, or can we sometimes stick to traditional methods? In that case, you can avoid some of the copyright issues altogether.

Copyright and other IP rights are not covered by the AI Act itself; there are separate, concrete laws for those areas.

Q. In what ways do you think AI is already making a difference in areas like patent and trademark searches while ensuring it is used responsibly?

A. A major advantage of using AI is certainly speed and efficiency. AI-powered tools can help find relevant results much faster, especially when it comes to patent or trademark searches. Accuracy could also improve in the long run, for example, when comparing logos or identifying the likelihood of confusion between trademarks. AI could take over certain administrative tasks that are still done manually today. However, I see these advantages more as a complement and support. In contentious decisions, human expertise will still be needed.

Q. We’ve talked about the benefits of AI, but it’s equally important to address the challenges. What do you see as the biggest risks when it comes to using AI in IP law? Are there any specific concerns that businesses and creators should be especially aware of?

A. A major risk in the field of copyright law is that AI-generated content does not qualify as an original creation. A certain level of originality is required for a work to be protected by copyright, and only a human can achieve that. AI generates outputs based on mathematical probabilities – that is not a creative act. Therefore, purely AI-generated works are not protected under copyright law.

In practice, this means no usage or exploitation rights can be granted for such content. Since copyright is not transferable, companies can easily find themselves in a legal grey area. I’m referring here to European law, where IP is harmonized across the EU. In other jurisdictions, such as the United States, the principles differ. There, economic exploitation rights are more important, which also allows for easier transferability.

In any case, it’s not so simple. For example, if I run an agency and sell a purely AI-generated logo to a client, I cannot grant any rights to it if I don’t have an original creative contribution. This opens new legal issues in client relationships that may not be covered by existing contracts or general terms and conditions. It is crucial to review this and seek legal advice by specialized lawfirms.

Liability for third-party copyright infringements is another issue since AI models are trained on data that may include copyright-protected material. Many AI service providers exclude liability in their terms and leave the responsibility to the users. Since, unfortunately, there is no global “copyright database” like there is for trademarks or patents (databases of trademark and patent offices), this risk cannot be fully eliminated – but it can at least be minimized.

Q. How have companies adapted their approach to AI and intellectual property in recent years? Have you noticed any major shifts in how businesses handle AI-related legal risks?

A. Companies are beginning to understand that the use of AI carries legal risks and is not a lawless space. As mentioned earlier, the AI literacy rules of Article 4 of the AI Act became binding in February 2025. Clear internal guidelines are needed to train employees and make them aware of which data may be entered into AI systems and which may not. It is also essential to understand that AI is not a search engine. It outputs mathematical probabilities, not facts. This means that particular caution is needed in legal contexts.

I would warmly advise not to draft contracts with AI if you are not a legal professional. You always need to ask: What is the status of the training data? Under which legal system was the AI trained? We are also familiar with the problem of hallucinations. If AI does not know the answer, it simply gives the most probable one. But probability is not the same as fact. And in legal matters, that can easily go wrong.

Q. AI and intellectual property have become a hot topic of debate. Can something created by AI actually be patented? And what about the AI systems themselves, do they qualify for any kind of patent protection?

A. No, AI itself cannot be patented. Algorithms or software are also not patentable unless they are part of a technical invention, such as a neural network integrated into a process. However, software can be protected by copyright law, namely, a work of literature. In IP law, the question often arises: If a specific right does not apply, is there perhaps a complementary protection regime?

There was recently a post on LinkedIn claiming that a prompt had been patented. That is not possible either. A prompt may qualify for copyright protection, but only if it shows a certain level of creativity and goes beyond what is usual. However, the practical question is: Where ist the benefit, if the prompt is protected but not the output it generates?

Q. In recent years, AI tools have become a part of everyday life, and one of the most common questions I hear is: If someone creates an image or text using AI, do they actually own it? Can AI-generated works be protected by copyright?

A. As I mentioned earlier, purely AI-generated output is not protected by copyright. This means that if you create a visual for social media and use it exactly as it is, someone else could simply take it. Whether that’s polite is another question – but from a legal perspective, it would be allowed.

It’s different, though, if the work is significantly modified by a human. In that case, the situation changes.

Or, for example, if an image is fused with a human-created work. In such a case, the overall work can be protected by copyright, but the AI-generated component on it´s own remains unprotected.

Q. There has been ongoing debate about whether AI could ever have legal rights like a person or a company. Why was this idea considered, and what were the main reasons for ultimately rejecting it?

A. The European Commission once discussed this possibility but ultimately rejected it. One possible reason is the issue of liability. An AI has no consciousness and therefore cannot make decisions with intent or negligence. Granting AI its own legal personality would only complicate things, especially in liability cases.

For now, it makes more sense to stick with existing legal concepts and adapt them to meet new challenges. Currently, there are only natural persons (humans) and legal persons (such as companies).

Q. Some of the topics we’re discussing today are also featured on your YouTube channel “the cherry compiler”. What inspired you to start this project, and what message do you hope to share with your audience?

A. I wanted to show that law is not an „old-fashioned“ or boring subject but a living field full of exciting topics. My goal was to explain complex legal matters in a way that is easy to understand and, most importantly, to give female inventors more visibility. My area of expertise, Intellectual Property, is still a niche. At Austrian universities, it’s not part of the regular curriculum-you only learn about it if you specialize in it. I always found that unfortunate.

The fact that IP is experiencing a real renaissance due to AI made the timing perfect to start this vlog!

Q. Everyone who watches “the cherry compiler” can find answers to many important legal questions—for example, can we use the image of the Mona Lisa in an AI-generated project? But for those who haven’t seen your channel yet, could you give us a brief answer here?

A. It depends – that’s the classic legal answer.

The original painting of the Mona Lisa is „public domain“ because its creator, Leonardo da Vinci, has been dead for several centuries. In Austria and Germany – the legal systems I refer to – works enter the status of public domain 70 years after the death of the creator. In the U.S., it’s 90 years. Public domain status means that the Mona Lisa’s image can generally be used freely.

However, it can be different with digital reproductions. E.g. some museums claim rights to high-resolution digital copies of their collection items, even if the original is public domain. They justify this by the effort and work involved in creating these reproductions.

In my vlog, I stick a moustache on a printed, physical image. That’s legally fine. But using a digital work as a base is more complicated, as reproduction rights may apply.

Q. Many people are excited about using AI, but legal risks can be a major concern. What practical advice would you give to anyone who wants to use AI while staying on the right side of copyright law?

A. Avoiding specific artist names in prompts is surely a good idea – no matter if they have already died or are still living. General styles, for example, are not protected by copyright. You could simply describe what the image should look like: “Create a floral image in the style of the early Baroque,” or something similar.

Companies should consider developing internal guidelines for the use of AI. It’s also worth reviewing contracts and setting clear rules for liability and quality assurance. This way, you can at least minimize the risks.

And in case of doubt: ask a specialized lawyer!

Q. And for our final question… With AI and new technologies advancing so quickly, what do you think the future of patent law will look like?

A. I think patent law is facing major challenges. With tools like improving AI-supported research, we will increasingly have to ask ourselves: Where does human innovation begin, and where is it just machine-driven advancement? In the future, it may become necessary to adapt current patent law to address these developments. In Europe, it’s also a question of ensuring that we don’t fall too far behind the U.S. and China due to regulations. Of course, the AI Act aims to build trust in AI, which is a noble goal. Still, we must be careful not to lag behind.

Another interesting question is whether entering an invention into an AI system could be considered as a disclosure that affects novelty. The rule is that an invention must not be made public before it is filed-like exhibiting it at a trade fair. Whether submitting an invention into an AI might theoretically undermine this concept is worth considering. It probably depends on how secure the system is and what happens to the entered data.