Open science and intellectual property are often framed as pulling in opposite directions: one built on free dissemination of knowledge, the other on protecting and rewarding innovation through exclusive rights. But as the pace of scientific publishing accelerates and generative AI reshapes how research is produced, the case for finding synergies between the two has never been more urgent.

This question sits at the heart of the upcoming STOA workshop “Rethinking intellectual property in the era of open science and AI”, taking place on October 13, 2026. The event in the European Parliament will bring together policymakers, researchers, and other stakeholders to explore flexible, cooperative approaches within existing intellectual property frameworks, and to consider what policy changes might be needed to help Europe's research reach its full potential.

Among the speakers is Gustav Nilsonne, associate professor of Neuroscience at the Karolinska Institute (Sweden), where he leads the metascience research team. Nilsonne is a long-standing advocate for open science and a work package leader in the Horizon Europe-funded IP4OS project, which aims to develop and disseminate knowledge on how intellectual property rights management can support open science.

Ahead of the workshop, we spoke with Nilsonne about why intellectual property matters for scientists, what a genuinely open science system might look like, and how AI is reshaping questions of authorship and ownership in research.

---

What does open science mean? If we talk about implementing “open science”, does that mean the current scientific system is “closed” – inaccessible, non transparent, elitist…?

Gustav Nilsonne: A lot of science is still not open, and I personally think that is a failure on the part of the research community. We write a lot of articles, and many of them end up behind paywalls, so they cannot be read by the general public and by journalists, and sometimes not even by other colleagues, because they don't work at a university with a subscription. It varies between different journals and different publishers, but oftentimes, journals are selling our papers back to other scientists.

In addition to that, there are also many parts of the research process that are not open. If you pick out a research paper, most likely the data underlying the study is not shared. Other digital objects used in the experiment or the study, such as the code used to analyse the data, are also not shared.

That's because we are still publishing papers the way they looked when everything was printed on paper. We have not yet managed the transition into a digital reality, and also not yet managed the transition into open science.

---

What would a fully open science system look like?

Gustav Nilsonne: Here we can imagine lots of things. A rather straightforward way to imagine it would be not having paywalls anymore – everything would be published with open access.

But we can also imagine the entire system of scientific communication being different. Instead of having all kinds of scientific outputs bundled together, perhaps in this world I could specialise as a scientist in generating data, and then someone else could do the analysis, and I could output various kinds of scientific value – scientific objects that are not necessarily in the form of text.

---

Why are intellectual property rights important for scientists?

Gustav Nilsonne: There are great expectations from society that we researchers will produce innovations, that we will make discoveries that can lead to patents, which can in turn lead to new products, new services, new companies, new technology, for helping people in all kinds of purposes.

To do that, we have to understand how copyright applies to our scientific outputs. Scientists produce text, data sets, images, code, and more: we have to understand how to navigate the rights around these research objects.

Patents are more relevant for researchers who make discoveries and inventions – in the natural sciences and medicine, for example. With patents, it's crucial to plan ahead and decide on the timing, so that you can first apply for a patent and then publish the results. Before we make a new invention, we have to know what we're going to do once we have it, so we can take it through the right channels.

It's very important that universities support their researchers here, because we cannot all be legal experts as well as scientists.

When it comes to open science, we often hear this phrase: “As open as possible, as closed as necessary”. But it needs interpretation: what is “as closed as necessary”? There's uncertainty among researchers – people are afraid of doing something wrong, and therefore they close their science more than necessary.

---

Your main field of research is “metascience”. Could you explain what metascience is, and why thinking about how science is produced and shared is relevant to questions of intellectual property?

Gustav Nilsonne: Metascience is research on research itself, using quantitative methods. We try to map the scientific process and measure how much transparency there is, how much reproducibility there is, in order to make research more useful and more trustworthy. We also do various kinds of interventions – we test ways to make research more open or more reliable.

It's excellent that we have policies for open science, but we also need to actually measure what happens once these policies are implemented and what the outcomes are. When we change processes, we need to test things and follow up on how they work. That's the sort of thing metascience does.

---

What's your take on AI being used to generate scientific hypotheses or help design experiments? Could ChatGPT be a co-author in a paper?

Gustav Nilsonne: It's an interesting question, and we've discussed this in the IP4OS project. Some people have suggested that AI models should be able to own copyrights, but personally I doubt it – the AI model is not a subject in law. In my opinion, there has to be a human who takes responsibility for the output they produce.

---

AI companies have scraped enormous volumes of published research to train models, without permission or compensation. How can researchers and universities get a share of that value?

Gustav Nilsonne: I think you are right about the training on scientific articles. It seems to me that the scientific literature, taken as a whole, is probably the most important training data these models have been trained on.

But my view is that I wrote my papers because I was hoping other people would read them and find them valuable. In this case, the AI companies have read them and found them valuable, and I think that's great. They have made innovations, and I am a happy user of AI models, so in that way the value has been returned to me. I don't see it as very problematic.