[00:00] Jonny Coates:
In recent months, researchers have watched data sets disappear from public repositories, governments restrict some international collaborations, and AI systems absorb vast amounts of scientific literature. These events all raise the same question: how open should science be? The future of science may depend less on choosing between openness and security, and more on understanding how trust makes both possible. Welcome to SIAcast, the podcast where we explore the science of science. I'm your host, Jonny Coates.
[00:29] Katharina Miller:
I am your co-host, Katharina Miller.
[00:32] Jonny Coates:
You'll also hear from
[00:33] María Bardají Cruz:
My name is María Bardají Cruz.
[00:34] Richard Gold:
I'm Richard Gold.
[00:35] Monica Granados:
Hi, my name is Monica Granados.
[00:37] Leslie McIntosh:
Hello, I'm Leslie McIntosh.
[00:42] Katharina Miller:
Open science is built on sharing, transparency, and international collaboration. Research security often introduces restrictions and controls. Are these goals inherently in tension, or can they reinforce one another?
[01:00] Richard Gold:
I think I'm gonna give you a lawyerly answer and say yes and no. There are certain aspects that are in tension, so if we're talking about defense industries, we're talking about attacks on sovereignty through, uh, cyberattacks and so on, where data could be changed, data could be taken, data could be misused. Uh, perhaps they're somewhat in tension. But I think fundamentally they're two aspects of the central problem of innovation. How do we create new knowledge out of old knowledge? We now recognize more and more – again, another change in society in the way we do innovation over the last 10 to 20 years, we've recognized that having access to large sets of data that are of high quality are necessary, and no one entity, no one country, can produce it by themselves. So there's a recognition we have to do it together, which includes necessarily some aspect of openness and of sharing of knowledge, which is fundamental to open science.
[02:05] Richard Gold:
It requires standardization across countries i- within the scientific and technology communities, again, not national, so that's part of open science. Transparency about how we collected the data so we can assess its quality and integrity. So all of that is necessary for innovation. One attack, uh, on sovereignty from a research integrity point of view is if someone attacks a data set, uh, database on a particular infrastructure and modifies it, deletes it, otherwise changes it. So in order to have open science, you actually need some level of cybersecurity. How do we make sure that bad actors can't go in and change the data or submit data of low quality that decreases the value of the entire data set? So I think open science needs some level of security in order to ensure that the project of open science goes ahead. On the other side, I think national security research integrity, uh, research, uh, sovereignty requires open science, because no one country wants to be left out.
[03:14] Richard Gold:
And if because of the international nature of the problem that we all have to contribute to the data, if we go it alone, we're essentially cutting ourselves off from what is going to feed innovation in the future. So in order to be sovereign, it means to not ever be dependent on someone else for advancing one's own research, and that's where open science contributes to, uh, research, uh, sovereignty, by ensuring that there are these open data sets that one could always access, this common knowledge through publications, through conferences, through dealing with people that will enable us to be able to use that knowledge. And us in that sense being, uh, a country or, or region. So I think OS needs some level of security, and security needs some level of open science.
[04:09] Jonny Coates:
And this is a really important point. Open science does not mean that everything must be open. Creative Commons produce licenses that enable researchers and others to effectively set conditions on how open a given work is.
[04:23] Monica Granados:
Yeah, I think it really encapsulates the, the purpose and the vision of Creative Commons when it was first created, which was about giving creators agency. And creators in this conversation, you know, it's gonna really mean researchers, scientists, like, knowledge creators. But we at Creative Commons also, uh, have creators who sit in, uh, the artistic realm or who are authors of books or photographers, curators of cultural works. And we wanted to give this wide spectrum of creators the ability to choose, to give them agency, so that the decision around whether something, you know, should be shared really belongs to the person who has created the material. And that's really embedded in, like, the structure of the licenses, you know, that they are copyright licenses. It's a way to give, uh, copyright holders the ability to give, uh, to have others reuse their work.
[05:33] Monica Granados:
So it's giving, uh, permissions to use what otherwise would be closed, protected, uh, all rights reserved copyright on their work. And so whether that's, you know, a picture or a publication or a data set, when you add that license, you're giving others the ability to use your work, and you're making that decision. And so I think the, you know, the data governance really comes into play as, like, an additional layer on top of the licenses. The licenses are just a tool to have others use the material. How it gets used, I think, is outside of the remit of the licenses. That really does belong in data governance because the, you know, the licenses don't have any terms around how the work gets used with, you know, with some exceptions around, like, you know, there's, um, certain provisions in the licenses, you know, example, where, like, an endorsement is not implied.
[06:30] Monica Granados:
There's specific, uh, you know, we have a spectrum of licenses that allow, uh, for use but with certain conditions or restrictions, so, you know, requiring attribution or a share-alike or a non-commercial or non-derivative. These are all elements of the spectrum of licenses. But beyond that, that is the role of, of data governance, and it's by design. That's... The licenses are just meant to share. What comes after that is data governance.
[06:58] Jonny Coates:
Licensing allows researchers to set limits on how their work is reused, but a key issue arises with public health data. This is where we see things tipping into data governance.
[07:08] Katharina Miller:
Health research depends increasingly on access to large data sets, yet health information is among the most sensitive personal data we have. How do we reconcile the goals of open science with the need to protect individual privacy?
[07:25] María Bardaji Cruz:
Katharina, I believe we need to, to take a little bit of a step back and reconsider what we mean with reconciliation, right? Because I try to look at this in a very, in a, in a slightly different way. So I believe that privacy and open science are not opposing forces, so they don't have to cancel each other out. I believe the real challenge is building more understanding and connection between both worlds to nurture trust and to let trust grow. And trust matters because without trust, patients will not be sharing data. Companies will not be sharing with each other, or public/private sectors will not collaborate, right? So we really need to understand that without trust, research, at the end of the day, will become weaker and less innovative. So maybe it's time to move past that idea that this is a trade-off, that this is either one or the other, right?
[08:27] María Bardají Cruz:
So data for research needs to be there and needs to be as open as possible, but it also needs to be protected, in my opinion, when it needs to be protected and where it needs to be protected, and this is nothing new, right? [laughs] So in foc- instead of focusing these from a conflict perspective, need to reconcile, need to, um, to find a solution to bring those matters together, I think it's rather labeling that tension but with a more positive word, so, and, and seeing this as an opportunity. An opportunity to show everybody, the society in general, that data can be used for research purposes in a responsible way, and this applies as well for, for AI. So w- we can do this in a thoughtful, in an ethical, and a trustworthy way, so both goals are in parallel.
[09:20] Katharina Miller:
Many researchers often rely on anonymization as one of those safeguards, and you al- already mentioned AI. And especially with advances in AI and data analytics, how confident can we still be that anonymized health data truly protects individuals?
[09:40] María Bardají Cruz:
It's a good question. [laughs] I would say I would also suggest moving away from that idea that anonymization is the only thing that we can implement because it's... and also it's not absolute, right? Anony- has never been. It's not just now with the explosion of AI capabilities and affordances that we see, oh, anonymization is not, not enough. It has never been, but it was never construed in that way. So I believe the GDPR also refers to reasona- reasonable likely means that can be used to re-identify, right? So it's an evolving concept. Also, if we look into for the particulars of research, we also see this tension between utility of the data and anonymization versus anonymization as a measure of data transformation, right? So that is a, a big, big challenge, and because the more we transform the data, the less usable is for research. So rather than looking at, or that having a binary view of anonymization, oh, it's anonymized, then we can do whatever we want, right?
[10:47] María Bardají Cruz:
We still see some comments, or we still see that assumption being present in our daily, um, professional discussions. What matters to me is the combination of anonymization, among all the things, but with other measures. So not seeing this as... Because it's a false sense of security, I believe that the idea that the data is anonymized, especially in the context of AI, right? So if the data is anonymized, I can do whatever I want. Privacy legislation is not lo- lo- no longer applicable. Well, I don't believe we are now in a position to still affirm and assume that that is fully realistic. What has changed with AI as a means to re-identify, I think is, is the scale and the sophistication of those means, right? We need to look beyond re-identification. We need to look into inference, into linkage. Because of course, modern m- models or a- analytics can... I mean, that's their purpose, right? To draw patterns, to identify tendencies, trends, patterns.
[12:00] María Bardají Cruz:
So that allows for- Inference of data, maybe not to re-identification. I believe the OECD mentions this in the, in their AI framework, which is even with anonymized data, there can be a privacy risk because of inference and linkage. So I think the short answer to the question would be anonymization is good, and we should keep on using it, of course, but it shouldn't be absolute. We should be looking at the context.
[12:30] Katharina Miller:
I'm very much involved lately in the world of open science advocates, and they often argue that sharing health data accelerates discoveries and improves healthcare outcomes. So María, where do you believe the limits of openness should be?
[12:47] María Bardají Cruz:
What does it mean sharing health data? What are we talking about? 'Cause I do not-- I cannot see open science advocates saying, "Let's share with the whole peop- you know, with the whole world. Let's put them in Internet, the record of a patient." Nobody's talking about that, so I d- I don't think that the, the, the question is whether we should have safeguards and limits. It's where we put those limits. So again, I think everything it's, it's related to the data should be as open as possible and as protected as necessary. That is the basis of every- everything. I don't see how a given research might need a full health record of a patient. So let's go back to all those preconceptions that we have about privacy, and let's revisit what do we mean with sharing health data. And also, let's, let's consider... Again, let's con- let, let's go back again to that discussion about anonymization. So if we want to share, how do you want to share?
[13:48] María Bardají Cruz:
How is, how is that advocacy for open science suggesting we need to share? Are we considering secure processing environments? Are we considering access controls? Are we considering retention periods, like I share with you for this given purpose, in this context, or in this particular environment to your two or three researchers within your institution for three months. Is that enough? Cool. After that, I recover the data. You have your results. You're good to go, right? But that level of previous analysis that we would need to do, those questions that we would need to ask ourselves are currently underestimated, I would say. So I would put them back in the table because otherwise, I mean, then I go back, it's, it's privacy by design again. It's considering a value, a right that needs to be protected without undermining another one, which is very valuable for society. So it's where we put the limits, not whether we have limits, I would say.
[14:52] Jonny Coates:
So far, we've talked about protecting people and protecting data. But for Leslie, discussions about research security often start in the wrong place. Research security isn't only about privacy. Increasingly, it's also about protecting knowledge itself.
[15:07] Leslie McIntosh:
So first of all, I wanna be clear on-- Well, I won't be clear, but at least acknowledge that research security has some very defined, uh, parameters, and it varies by country, and it has to or it tends to do with whether there is a risk, an economic risk, an IP risk, certain risks to the university, to the country, perhaps to the individual, depending on, uh, how it is. When I think about integrity and when I think about security, I think about, first of all, the integrity of science as a whole, so the integral of how are we looking at how do we trust and secure our knowledge more than a person or an IP or economics. I'm thinking more about the security of the knowledge and the tr-- and the trustworthiness of that. So it's a little bit of a, an addition to some other research security. So I just wanna put that out there. I come from a biomedical background. I come from a background of overseeing the aggregation of biomedical data, of clinical data.
[16:23] Leslie McIntosh:
This is back in the day when it was, it was very new. And so we had to bake in security, monitoring some potential, you know, preventing breaches, things like that, right into that. So it's been what I've done for and thought about for a while. Now, I want to say that I may be com- coming from biomedical, but I think that if we're not thinking about the trustworthiness of knowledge in a holistic sense, then we're really missing something, and that cannot exclude the humanities and social sciences because they're critically important to our environment.
[17:00] Jonny Coates:
What do you think some of the biggest misconceptions are around research security?
[17:04] Leslie McIntosh:
I think there's a few things going on. One is, is this gonna make my researcher research harder to do? Because there's so much as-- administrative burden that's put on researchers to get done that takes them away from the actual content of what they're trying to look at. And quite frankly, I can understand that resistance. And, and I don't think that we should be putting the responsibility of this onto people who need to get back to the research at hand and not be s- not have to think about whether their research could be misused or how it could be, and to put in all of those preventative measures that you can. I, I do think there needs to be support around that. For many people, certainly not all in the world by any stretch of the imagination, since the wall came down in 1989 The idea, I think there's been this gestalt that the world is moving towards democracy, democratizing everything, and that the world is good, and we can trust people, we can travel.
[18:18] Leslie McIntosh:
And the fact is that there's geopolitical tensions that have been there. The world hasn't been a safe place for everybody during that time, not in the Balkans, not in Darfur, not in many places. We can cite many places that haven't been secure during that time. But for many people in the Global North, this has been a time of relaxation from the warming from the Cold War, if you will, and it's just not the case. There... It is the fact that there are countries, there are entities within countries, there's ideologies that are using science and scientists to manipulate their, the world into the view that they want, and not pointing out to... not paying attention to the scientific question. Where that comes into scienti- to researchers itself is, how do you differentiate between the person that I know and the country that they are from? And how do you... Do you trust the person even though you don't trust the country?
[19:32] Leslie McIntosh:
Do you have to perhaps bring in some reticence because of people from certain countries? And that makes people very, some people very uncomfortable, and it should. I mean, these are, these are hard questions. Again, gonna go back to the, the social sciences. They actually study these things and understand a bit.
[19:53] Jonny Coates:
Why do you think academia is so resistant to embracing these kind of changes? Why-
[19:58] Leslie McIntosh:
Well, I think that it, the system worked for some people. Uh, honestly, it worked for the privileged very well, that you could work in this trusted ecosystem. I think there was a lot of problems with it, and I think there's still some problems with it that all of the games that are being played now are not necessarily addressing, and I'm particularly thinking about the inequities of people having access to education and research, and the ability to get jobs depending on their background, their country, their gender, all of the other things. So why do I think that academia, though, is resistant? Let's just stop there. Academia is resistant, period. It was resistant to open science, and then open science became bigger, and the pandemic, the most recent pandemic, showed us that there are some really good things about open, having open data, and the ability to also have preprints so that you could rapidly more publicly exchange knowledge.
[21:11] Leslie McIntosh:
But when you do that, the people consuming that knowledge do not have the background or the training necessary to understand the nuances. And it's nothing against them. I just got off of a plane, and I expect it to work and the pilots to know what they're doing without me knowing, except for that I get on the plane, show my ID, you know, sit down, and I want everything to work. I don't think that's wrong. The problem is when people take this information and assume that they know as much as the experts and want to use the information in a non-nuan- in a generalized way when it's actually very specific information. And so there's a disconnect between availability and the ability to understand what is available.
[22:02] Katharina Miller:
Many governments are becoming more concerned about the strategic value of research in areas such as AI, biotechnology, quantum technologies, and advanced materials. What do you think? How should universities balance scientific openness with national and economic security concerns?
[22:23] Richard Gold:
Again, I think openness is critical to the scientific process, especially in many of the areas you just mentioned. They're moving exceedingly quickly. If I try to close it off, A, it requires, uh, me to build my own servers, or I'm locked out. And so how am I gonna get the latest knowledge? How do I promote what I'm doing as an academic if I can't go to international conferences and release the results of what I'm doing? So I can be second, third, fourth-tier researcher in my little corner, or I can be a leading player. And if our researchers aren't leading players, neither are our firms. Our firms depend on their relationships with researchers. So it's both in the f- private sector's interest and the academic interest to be open. And then from a national security point of view, again, if our... if we don't have people at the center of the discussions around AI, around quantum, around biotechnology, then we don't have the capacity to know where the front...
[23:27] Richard Gold:
to, to have knowledge of the frontier and be able to use that frontier of knowledge nationally. And so there's really no choice. Or there is a choice. There is a choice of being open and transparent and collaborative and being at the cutting edge.
[23:41] Katharina Miller:
So we attended the World Conference on Research Integrity this year in Vancouver, and I have to say I was very astonished that, um, many of the topics were focused on research security, and you were also invited as a keynote speaker. And being in this c- at this conference, uh, we could listen to many colleagues from worldwide and their increasing restrictions on international collaborations in certain scientific fields. What risks do you see if research security policies become too restrictive?
[24:17] Richard Gold:
So if countries, in my view, wrongly interpret research security as going your own direction and not sharing, uh, then it risks removing your scientists from, you know, the leading edge of research, especially in fast-moving fields like AI, quantum, uh, some areas of biotechnology, et cetera, materials as you mentioned. And so you're taking people out of the game, which means that if your country has... you know, needs to develop a technology, that they're not going to have the internal resources to do so. So at some point they'll be forced to look externally, exposing themselves to the very risk that they were concerned about. On the other hand, there are some issues, um, at, at-- that I'm more ambivalent about, right? So we have leading researchers in AI saying, "Look, we're, we're at...
[25:12] Richard Gold:
The frontier of AI is now dangerous, not just to states, but to humanity, and we need to slow down." But just to say, you know, we're, we're not gonna allow research in certain areas, I think, um, will undermine the flexibility of the innovation system. We never predict well where science is going to go, where technology's going to go. The history, you know, even the head of IBM doubted when they developed a personal computer that it would change much. Nobody, you know, the people talking about television when it first came out said, "Well, no one's gonna sit down on a couch and watch this thing except for rare occasions. They're gonna be moving about, so radio is going to, to live." Basically, all of our predictions about what humans will do with the technology have been wrong.
[26:06] Richard Gold:
That doesn't mean we shouldn't have systems to monitor it, that we should not anticipate possible negative outcomes and have a regulatory regime that would be alerted if these certain lines are crossed and take action. So it's not a, an excuse for inaction, but just to close down research in the name of research sovereignty, to me seems the wrong approach.
[26:29] Jonny Coates:
And I mean, that kind of leads us into talking about, not explicitly the situation in the US, but somewhat. So we've seen a lot of datasets disappear, public repositories, government websites, and so on. Regardless of what the reason for that disappearance may be, could you talk about any potential protection that CC licenses offer against losing those kind of things?
[26:51] Monica Granados:
Yeah. Yeah, it's a little bit sort of secondary. So a benefit of having any type of work be CC licensed is that the licenses are, uh, irrevocable. So that means if I create a dataset and I put a CC license on it, and then someone downloads that CC licensed dataset, I can never take the permissions that I have granted to that reuser away. So in this very, in this specific case of sort of datasets that are disappearing or, you know, are let's just say are no longer publicly accessible, if that original dataset has a CC license on it, it really gives you the permission to be able to have, uh, copies of those datasets in other repositories, because the license gives you that explicit permission, and that permission, again, is irrevocable.
[27:55] Jonny Coates:
So while CC licenses are great at protecting datasets and combating some of the geopolitics, what about AI and the impact of large-scale data mining? This has become a very contentious topic recently.
[28:06] Richard Gold:
Well, my whole research history shows that I believe openness is the way to go, and that closed knowledge is harmful. So I don't think it ever made sense, uh, to have too big a distinction. I think it, you know, it might go along with the stage of development, right? So closed knowledge about a particular application of, in, in a particular technology to provide exclusivity so that one has exclusivity, I can understand, because it doesn't really foreclose someone else from developing a competing product or better product, et cetera. Uh, but the, this fundamental... And I don't know if we've ever had a truly completely closed system. I think it's idealized as well. Lots of knowledge flows, uh, people talk, e-e-et cetera. But I think with the needs for large datasets to train AI and to validate the AI, it becomes harder to justify closed systems, because closed systems means idiosyncratic systems collected differently.
[29:12] Richard Gold:
It might be a high-quality dataset, but it's small, and to make it valuable, it needs to work with other datasets, and it's more or less impossible to do that if we have a bunch of different silos of closed knowledge. We need to open it up. So I think AI in general, uh, not just generative AI, but all the various forms of AI, are pushing us for, to large datasets, which naturally has an open aspect to it. And then the debate moves further down. At what point do we stop sharing? There's a, a strong argument for open source software, uh, behind the AI. What about the models and the weights and all the other aspects that go into an AI? And so I think that's where the debate is. Where's the dividing line? What is a particular application of knowledge that is best advanced through some form of exclusivity, and what needs to be open to keep the whole industry and research enterprise going? I don't think we have an answer on that yet.
[30:19] Jonny Coates:
Maybe could we talk a little bit from the stepping back- Place to begin with, and talk about how these have been trained on, on, um, a lot of open data, a lot of scientific literature in particular, and, and sort of what Creative Commons thinks about that. Um, is that a thing you feel was-- Do you think it changes what we think and mean by open going forward, or do you think actually things are all working as, as they were expected, this is just maybe a slightly cheeky use?
[30:51] Monica Granados:
Yeah. That's a great question, and a question that is really central to a lot of the conversations we're having at Creative Commons. I think it's incumbent on me to recognize what an impact the adoption and rise and sort of availability of large language models, commercially available models has had on the work that Creative Commons is doing. So CC licenses are copyright licenses, and in many jurisdictions and for many applications, large language models fall outside of copyright. And what that means is that it really puts into question whether or not the licenses are invoked in the use of, of CC license material for training. And so from CC's perspective, you know, the specific question around whether the CC licenses imply that their content is, you know, that there's some permission that is being given for training, from CC's perspective, it's that the licenses do not necessarily imply that the content is intended by the licensor to be used for AI training.
[32:12] Monica Granados:
It was, you know, for humans to use, but for new technologies to use it as, as well. And so I think what's changing for, for CC, and to get to your question about like, about open, is that I think in the context of AI, the question about open is now really fragmented by sector because of the, the scale at which AI is ingesting material, and ultimately about the lack of reciprocity in the system. So it wasn't a problem-- And, you know, these are, these are generalizations, but creators like artists or photographers who were putting things online, but their work wasn't being ingested at scale, it was only sort of being re-shared by humans. There was a, there was not this like outsized effect of their, of, of a potential, um, you know, jeopardizing, like je- jeopardizing their, their business. And getting nothing in return has a pretty big consequence for creators who rely on these types of, you know, on, on the, on their work to make a living.
[33:29] Monica Granados:
And on, you know, the other side of the spectrum, like researchers have a mandate to continue to share their work. They're adopting a lot of, um, AI workflows, adopting agentic, uh, or agents in their, in their workflows, so this is just sort of a part of how they do research now. I think there's a request for some reciprocity in the, in the form of, of attribution for researchers, but there's definitely just different relationships with AI from a body of creators who used to be more homogeneous when they were using the licenses, and now they're very distinct. And I think that does kind of change the question of open because now there's this sort of element of open may, may depend, or the definition of open might be different for different sectors.
[34:28] Jonny Coates:
So let's talk about some of those solutions. So s- signals is sort of the most visible thing I would say that you've, you've done around, directly around AI. So could you talk a little bit about what the signals are, what they in- what they were intended for?
[34:41] Monica Granados:
Yeah. It really is, uh, you know, a response to this lack of reciprocity in a system. The, the-- and the recognition that the licenses m- may not be fit for purpose it, for purpose for AI use, so for, for, for machine reuse. And so, you know, we started to think about, um, you know, first surfacing this like lack of reciprocity and, and really importantly, thinking about and challenging us and the community to think about what's the consequence to the commons if we don't intervene, both as CC and as, you know, uh, uh, the, the open community who is stewarding the commons. And so about a year ago, we produced a, a, a report that really spoke to this, the lack of reciprocity, what's gonna happen to the commons? Are people going to stop sharing? [upbeat music] We're gonna be issuing, um, updated guidance around how the licenses do or do not interact with AI and surfacing how that may be different in different jurisdictions, because different jurisdictions have specific provisions for AI.
[36:08] Monica Granados:
Certain jurisdictions have fair use or fair dealings and others do not. So this guidance, we hope, will really answer a lot of questions around how the CC licenses operate in an age of AI, because we are anecdotally starting to see that more researchers, authors ... are selecting more restrictive licenses in an effort to prevent AI from ingesting their material when all that action does is make it harder for humans to reuse their material. So that's our first intervention. Our second intervention is around making attribution the norm in AI. So, you know, Jonny, you, uh, very astutely pointed out that what researchers want is credit. They want attribution. That's the type of reciprocity that they want back for the, you know, AI using their work. They want credit for the, for the knowledge they created. So this work is thinking about how can we enable attribution in machine use or in AI systems, where is that being done?
[37:17] Jonny Coates:
Licenses are one way of providing some research security and in navigating AI. Monica mentioned earlier about governance and where the responsibility of CC ended, and as Richard alluded to, companies can decide to not be open with their data. But I want to broaden this, and so I asked Leslie about who is responsible for research security.
[37:36] Leslie McIntosh:
So part of it is each country's government is responsible for the security because they also have an economic obligation to their citizens to protect and to potentially, hopefully, not all countries do this, but to make their countries economically viable and sustainable for their people, particularly if you're talking about a democracy. I'm very much pro-democracy. So there is that aspect. I think also, and this is what part of forensic scientometrics is about, is I think there is a place for a group that's outside of a university or a company or a government or an institution, so like an NGO type, to both monitor and... I would just say monitor and help perhaps influence policy when it comes to that and when it comes to research security. So besides government, who have obviously economic interest in that, institutions have, have an obligation to this.
[38:46] Leslie McIntosh:
Right now, I feel like probably because it's the easiest to look at and it's the most exposed, one might think that publishers have a large responsibility in here. I am not saying that they are responsible free, but I think that the onus is other, is, is in another place. But as far as the responsibility, I'm gonna put it more at the institutional level rather than the individual, rather than the researcher itself. That being said, there are certain research fields that are very sensitive, that are more known to be sensitive, some that you wouldn't think of necessarily. The health field we've already talked about, and there's certain laws about that, but most of the laws have to do with protecting individual rights rather than a community. Other places that aren't biomedical, you talk to anyone who works with rare minerals, and they don't necessarily wanna share their data because then it can... they... those places might go and be, uh, destroyed, robbed. There's other risks to it.
[39:53] Leslie McIntosh:
Uh, anthropologists who study in particular individuals don't want their whereabouts known. So there's, there's lots of reasons that researchers might ha- want something to do with security, but I don't think it's security in the sense that most of the mandates are right now, which is about dual use and disclosing multiple affiliations and things, things like that.
[40:19] Katharina Miller:
So what practical steps would you recommend to a university now, today, to balance open science commitments with privacy and security obligations?
[40:32] María Bardají Cruz:
I think in sum, I would say, I will keep it very simple, be responsible. It's not that complex. I mean, it, it, of course it requires an effort, but if you consult the appropriate professionals and, you know, include that knowledge within your organization to help you understand how to use personal data in a responsible way, I think it would be easier for institutions to comply. So I would say certain measures, for example. So focus on purpose. What are you going to, to share the data for? Is it scientifically, is it, is it scientifically valuable? It is going to bring a value to society? Does it merit the access to the data? Keep control. Again, control. Use secure processing environments. Why not? Allow that access, but in a very particular setup with controlled environment, with secure environment. Think about, again, the life cycle, not just the access.
[41:35] Jonny Coates:
To, to round this out, looking ahead sort of five, 10 years' time, what do you think the relationship between openness, privacy, and security in science and academia will look like?
[41:45] Leslie McIntosh:
Oh, I wish I could say that it would be that idealistic picture where, where institutions have the resources for their researchers so that they could best share knowledge while adding trust to it. I'm not gonna say that's gonna happen in five years. I think five years people are gonna be complying with whatever security mandates are in place, and there's gonna be a lot of leakage all around. I do think, though, in 10 years that there might be less naiveté when it comes to the fact that the world was maybe as open as it was gonna get, and that we need to pull back and understand how this looks open and trusted simultaneously. And I don't think that you have to give up one without the other. I just think that we need to acknowledge it. And I know I've brought in a lot of things in here, but I also think about, like, an open river.
[42:44] Leslie McIntosh:
If you keep throwing junk into it and don't have any protections or regulations for not throwing junk in it, it's going to keep being polluted, and it's gonna be polluted even if you have those regulations. Then you have cleanup, and we need to do that when it comes to research to keep things open, available with understanding the acknowledgment that there is privacy, but privacy could also be used as an excuse for not having things open, and that the security of science, I'd like to think about that we move towards knowledge security and thinking about that from the academic space. But there will and are mandates coming through. The European Commission is working on a new one. Um, other countries are as well. That's not going away.
[43:35] Jonny Coates:
Open science and research security are often presented as opposing forces, but every conversation in this episode pointed somewhere else. The real challenge isn't deciding how much openness or security we want. It's deciding how we build trust, trust in data, trust in institutions, trust in collaborators, and ultimately trust in knowledge itself.
In recent months, researchers have watched data sets disappear from public repositories, governments restrict some international collaborations, and AI systems absorb vast amounts of scientific literature. These events all raise the same question: how open should science be? The future of science may depend less on choosing between openness and security, and more on understanding how trust makes both possible. Welcome to SIAcast, the podcast where we explore the science of science. I'm your host, Jonny Coates.
[00:29] Katharina Miller:
I am your co-host, Katharina Miller.
[00:32] Jonny Coates:
You'll also hear from
[00:33] María Bardají Cruz:
My name is María Bardají Cruz.
[00:34] Richard Gold:
I'm Richard Gold.
[00:35] Monica Granados:
Hi, my name is Monica Granados.
[00:37] Leslie McIntosh:
Hello, I'm Leslie McIntosh.
[00:42] Katharina Miller:
Open science is built on sharing, transparency, and international collaboration. Research security often introduces restrictions and controls. Are these goals inherently in tension, or can they reinforce one another?
[01:00] Richard Gold:
I think I'm gonna give you a lawyerly answer and say yes and no. There are certain aspects that are in tension, so if we're talking about defense industries, we're talking about attacks on sovereignty through, uh, cyberattacks and so on, where data could be changed, data could be taken, data could be misused. Uh, perhaps they're somewhat in tension. But I think fundamentally they're two aspects of the central problem of innovation. How do we create new knowledge out of old knowledge? We now recognize more and more – again, another change in society in the way we do innovation over the last 10 to 20 years, we've recognized that having access to large sets of data that are of high quality are necessary, and no one entity, no one country, can produce it by themselves. So there's a recognition we have to do it together, which includes necessarily some aspect of openness and of sharing of knowledge, which is fundamental to open science.
[02:05] Richard Gold:
It requires standardization across countries i- within the scientific and technology communities, again, not national, so that's part of open science. Transparency about how we collected the data so we can assess its quality and integrity. So all of that is necessary for innovation. One attack, uh, on sovereignty from a research integrity point of view is if someone attacks a data set, uh, database on a particular infrastructure and modifies it, deletes it, otherwise changes it. So in order to have open science, you actually need some level of cybersecurity. How do we make sure that bad actors can't go in and change the data or submit data of low quality that decreases the value of the entire data set? So I think open science needs some level of security in order to ensure that the project of open science goes ahead. On the other side, I think national security research integrity, uh, research, uh, sovereignty requires open science, because no one country wants to be left out.
[03:14] Richard Gold:
And if because of the international nature of the problem that we all have to contribute to the data, if we go it alone, we're essentially cutting ourselves off from what is going to feed innovation in the future. So in order to be sovereign, it means to not ever be dependent on someone else for advancing one's own research, and that's where open science contributes to, uh, research, uh, sovereignty, by ensuring that there are these open data sets that one could always access, this common knowledge through publications, through conferences, through dealing with people that will enable us to be able to use that knowledge. And us in that sense being, uh, a country or, or region. So I think OS needs some level of security, and security needs some level of open science.
[04:09] Jonny Coates:
And this is a really important point. Open science does not mean that everything must be open. Creative Commons produce licenses that enable researchers and others to effectively set conditions on how open a given work is.
[04:23] Monica Granados:
Yeah, I think it really encapsulates the, the purpose and the vision of Creative Commons when it was first created, which was about giving creators agency. And creators in this conversation, you know, it's gonna really mean researchers, scientists, like, knowledge creators. But we at Creative Commons also, uh, have creators who sit in, uh, the artistic realm or who are authors of books or photographers, curators of cultural works. And we wanted to give this wide spectrum of creators the ability to choose, to give them agency, so that the decision around whether something, you know, should be shared really belongs to the person who has created the material. And that's really embedded in, like, the structure of the licenses, you know, that they are copyright licenses. It's a way to give, uh, copyright holders the ability to give, uh, to have others reuse their work.
[05:33] Monica Granados:
So it's giving, uh, permissions to use what otherwise would be closed, protected, uh, all rights reserved copyright on their work. And so whether that's, you know, a picture or a publication or a data set, when you add that license, you're giving others the ability to use your work, and you're making that decision. And so I think the, you know, the data governance really comes into play as, like, an additional layer on top of the licenses. The licenses are just a tool to have others use the material. How it gets used, I think, is outside of the remit of the licenses. That really does belong in data governance because the, you know, the licenses don't have any terms around how the work gets used with, you know, with some exceptions around, like, you know, there's, um, certain provisions in the licenses, you know, example, where, like, an endorsement is not implied.
[06:30] Monica Granados:
There's specific, uh, you know, we have a spectrum of licenses that allow, uh, for use but with certain conditions or restrictions, so, you know, requiring attribution or a share-alike or a non-commercial or non-derivative. These are all elements of the spectrum of licenses. But beyond that, that is the role of, of data governance, and it's by design. That's... The licenses are just meant to share. What comes after that is data governance.
[06:58] Jonny Coates:
Licensing allows researchers to set limits on how their work is reused, but a key issue arises with public health data. This is where we see things tipping into data governance.
[07:08] Katharina Miller:
Health research depends increasingly on access to large data sets, yet health information is among the most sensitive personal data we have. How do we reconcile the goals of open science with the need to protect individual privacy?
[07:25] María Bardaji Cruz:
Katharina, I believe we need to, to take a little bit of a step back and reconsider what we mean with reconciliation, right? Because I try to look at this in a very, in a, in a slightly different way. So I believe that privacy and open science are not opposing forces, so they don't have to cancel each other out. I believe the real challenge is building more understanding and connection between both worlds to nurture trust and to let trust grow. And trust matters because without trust, patients will not be sharing data. Companies will not be sharing with each other, or public/private sectors will not collaborate, right? So we really need to understand that without trust, research, at the end of the day, will become weaker and less innovative. So maybe it's time to move past that idea that this is a trade-off, that this is either one or the other, right?
[08:27] María Bardají Cruz:
So data for research needs to be there and needs to be as open as possible, but it also needs to be protected, in my opinion, when it needs to be protected and where it needs to be protected, and this is nothing new, right? [laughs] So in foc- instead of focusing these from a conflict perspective, need to reconcile, need to, um, to find a solution to bring those matters together, I think it's rather labeling that tension but with a more positive word, so, and, and seeing this as an opportunity. An opportunity to show everybody, the society in general, that data can be used for research purposes in a responsible way, and this applies as well for, for AI. So w- we can do this in a thoughtful, in an ethical, and a trustworthy way, so both goals are in parallel.
[09:20] Katharina Miller:
Many researchers often rely on anonymization as one of those safeguards, and you al- already mentioned AI. And especially with advances in AI and data analytics, how confident can we still be that anonymized health data truly protects individuals?
[09:40] María Bardají Cruz:
It's a good question. [laughs] I would say I would also suggest moving away from that idea that anonymization is the only thing that we can implement because it's... and also it's not absolute, right? Anony- has never been. It's not just now with the explosion of AI capabilities and affordances that we see, oh, anonymization is not, not enough. It has never been, but it was never construed in that way. So I believe the GDPR also refers to reasona- reasonable likely means that can be used to re-identify, right? So it's an evolving concept. Also, if we look into for the particulars of research, we also see this tension between utility of the data and anonymization versus anonymization as a measure of data transformation, right? So that is a, a big, big challenge, and because the more we transform the data, the less usable is for research. So rather than looking at, or that having a binary view of anonymization, oh, it's anonymized, then we can do whatever we want, right?
[10:47] María Bardají Cruz:
We still see some comments, or we still see that assumption being present in our daily, um, professional discussions. What matters to me is the combination of anonymization, among all the things, but with other measures. So not seeing this as... Because it's a false sense of security, I believe that the idea that the data is anonymized, especially in the context of AI, right? So if the data is anonymized, I can do whatever I want. Privacy legislation is not lo- lo- no longer applicable. Well, I don't believe we are now in a position to still affirm and assume that that is fully realistic. What has changed with AI as a means to re-identify, I think is, is the scale and the sophistication of those means, right? We need to look beyond re-identification. We need to look into inference, into linkage. Because of course, modern m- models or a- analytics can... I mean, that's their purpose, right? To draw patterns, to identify tendencies, trends, patterns.
[12:00] María Bardají Cruz:
So that allows for- Inference of data, maybe not to re-identification. I believe the OECD mentions this in the, in their AI framework, which is even with anonymized data, there can be a privacy risk because of inference and linkage. So I think the short answer to the question would be anonymization is good, and we should keep on using it, of course, but it shouldn't be absolute. We should be looking at the context.
[12:30] Katharina Miller:
I'm very much involved lately in the world of open science advocates, and they often argue that sharing health data accelerates discoveries and improves healthcare outcomes. So María, where do you believe the limits of openness should be?
[12:47] María Bardají Cruz:
What does it mean sharing health data? What are we talking about? 'Cause I do not-- I cannot see open science advocates saying, "Let's share with the whole peop- you know, with the whole world. Let's put them in Internet, the record of a patient." Nobody's talking about that, so I d- I don't think that the, the, the question is whether we should have safeguards and limits. It's where we put those limits. So again, I think everything it's, it's related to the data should be as open as possible and as protected as necessary. That is the basis of every- everything. I don't see how a given research might need a full health record of a patient. So let's go back to all those preconceptions that we have about privacy, and let's revisit what do we mean with sharing health data. And also, let's, let's consider... Again, let's con- let, let's go back again to that discussion about anonymization. So if we want to share, how do you want to share?
[13:48] María Bardají Cruz:
How is, how is that advocacy for open science suggesting we need to share? Are we considering secure processing environments? Are we considering access controls? Are we considering retention periods, like I share with you for this given purpose, in this context, or in this particular environment to your two or three researchers within your institution for three months. Is that enough? Cool. After that, I recover the data. You have your results. You're good to go, right? But that level of previous analysis that we would need to do, those questions that we would need to ask ourselves are currently underestimated, I would say. So I would put them back in the table because otherwise, I mean, then I go back, it's, it's privacy by design again. It's considering a value, a right that needs to be protected without undermining another one, which is very valuable for society. So it's where we put the limits, not whether we have limits, I would say.
[14:52] Jonny Coates:
So far, we've talked about protecting people and protecting data. But for Leslie, discussions about research security often start in the wrong place. Research security isn't only about privacy. Increasingly, it's also about protecting knowledge itself.
[15:07] Leslie McIntosh:
So first of all, I wanna be clear on-- Well, I won't be clear, but at least acknowledge that research security has some very defined, uh, parameters, and it varies by country, and it has to or it tends to do with whether there is a risk, an economic risk, an IP risk, certain risks to the university, to the country, perhaps to the individual, depending on, uh, how it is. When I think about integrity and when I think about security, I think about, first of all, the integrity of science as a whole, so the integral of how are we looking at how do we trust and secure our knowledge more than a person or an IP or economics. I'm thinking more about the security of the knowledge and the tr-- and the trustworthiness of that. So it's a little bit of a, an addition to some other research security. So I just wanna put that out there. I come from a biomedical background. I come from a background of overseeing the aggregation of biomedical data, of clinical data.
[16:23] Leslie McIntosh:
This is back in the day when it was, it was very new. And so we had to bake in security, monitoring some potential, you know, preventing breaches, things like that, right into that. So it's been what I've done for and thought about for a while. Now, I want to say that I may be com- coming from biomedical, but I think that if we're not thinking about the trustworthiness of knowledge in a holistic sense, then we're really missing something, and that cannot exclude the humanities and social sciences because they're critically important to our environment.
[17:00] Jonny Coates:
What do you think some of the biggest misconceptions are around research security?
[17:04] Leslie McIntosh:
I think there's a few things going on. One is, is this gonna make my researcher research harder to do? Because there's so much as-- administrative burden that's put on researchers to get done that takes them away from the actual content of what they're trying to look at. And quite frankly, I can understand that resistance. And, and I don't think that we should be putting the responsibility of this onto people who need to get back to the research at hand and not be s- not have to think about whether their research could be misused or how it could be, and to put in all of those preventative measures that you can. I, I do think there needs to be support around that. For many people, certainly not all in the world by any stretch of the imagination, since the wall came down in 1989 The idea, I think there's been this gestalt that the world is moving towards democracy, democratizing everything, and that the world is good, and we can trust people, we can travel.
[18:18] Leslie McIntosh:
And the fact is that there's geopolitical tensions that have been there. The world hasn't been a safe place for everybody during that time, not in the Balkans, not in Darfur, not in many places. We can cite many places that haven't been secure during that time. But for many people in the Global North, this has been a time of relaxation from the warming from the Cold War, if you will, and it's just not the case. There... It is the fact that there are countries, there are entities within countries, there's ideologies that are using science and scientists to manipulate their, the world into the view that they want, and not pointing out to... not paying attention to the scientific question. Where that comes into scienti- to researchers itself is, how do you differentiate between the person that I know and the country that they are from? And how do you... Do you trust the person even though you don't trust the country?
[19:32] Leslie McIntosh:
Do you have to perhaps bring in some reticence because of people from certain countries? And that makes people very, some people very uncomfortable, and it should. I mean, these are, these are hard questions. Again, gonna go back to the, the social sciences. They actually study these things and understand a bit.
[19:53] Jonny Coates:
Why do you think academia is so resistant to embracing these kind of changes? Why-
[19:58] Leslie McIntosh:
Well, I think that it, the system worked for some people. Uh, honestly, it worked for the privileged very well, that you could work in this trusted ecosystem. I think there was a lot of problems with it, and I think there's still some problems with it that all of the games that are being played now are not necessarily addressing, and I'm particularly thinking about the inequities of people having access to education and research, and the ability to get jobs depending on their background, their country, their gender, all of the other things. So why do I think that academia, though, is resistant? Let's just stop there. Academia is resistant, period. It was resistant to open science, and then open science became bigger, and the pandemic, the most recent pandemic, showed us that there are some really good things about open, having open data, and the ability to also have preprints so that you could rapidly more publicly exchange knowledge.
[21:11] Leslie McIntosh:
But when you do that, the people consuming that knowledge do not have the background or the training necessary to understand the nuances. And it's nothing against them. I just got off of a plane, and I expect it to work and the pilots to know what they're doing without me knowing, except for that I get on the plane, show my ID, you know, sit down, and I want everything to work. I don't think that's wrong. The problem is when people take this information and assume that they know as much as the experts and want to use the information in a non-nuan- in a generalized way when it's actually very specific information. And so there's a disconnect between availability and the ability to understand what is available.
[22:02] Katharina Miller:
Many governments are becoming more concerned about the strategic value of research in areas such as AI, biotechnology, quantum technologies, and advanced materials. What do you think? How should universities balance scientific openness with national and economic security concerns?
[22:23] Richard Gold:
Again, I think openness is critical to the scientific process, especially in many of the areas you just mentioned. They're moving exceedingly quickly. If I try to close it off, A, it requires, uh, me to build my own servers, or I'm locked out. And so how am I gonna get the latest knowledge? How do I promote what I'm doing as an academic if I can't go to international conferences and release the results of what I'm doing? So I can be second, third, fourth-tier researcher in my little corner, or I can be a leading player. And if our researchers aren't leading players, neither are our firms. Our firms depend on their relationships with researchers. So it's both in the f- private sector's interest and the academic interest to be open. And then from a national security point of view, again, if our... if we don't have people at the center of the discussions around AI, around quantum, around biotechnology, then we don't have the capacity to know where the front...
[23:27] Richard Gold:
to, to have knowledge of the frontier and be able to use that frontier of knowledge nationally. And so there's really no choice. Or there is a choice. There is a choice of being open and transparent and collaborative and being at the cutting edge.
[23:41] Katharina Miller:
So we attended the World Conference on Research Integrity this year in Vancouver, and I have to say I was very astonished that, um, many of the topics were focused on research security, and you were also invited as a keynote speaker. And being in this c- at this conference, uh, we could listen to many colleagues from worldwide and their increasing restrictions on international collaborations in certain scientific fields. What risks do you see if research security policies become too restrictive?
[24:17] Richard Gold:
So if countries, in my view, wrongly interpret research security as going your own direction and not sharing, uh, then it risks removing your scientists from, you know, the leading edge of research, especially in fast-moving fields like AI, quantum, uh, some areas of biotechnology, et cetera, materials as you mentioned. And so you're taking people out of the game, which means that if your country has... you know, needs to develop a technology, that they're not going to have the internal resources to do so. So at some point they'll be forced to look externally, exposing themselves to the very risk that they were concerned about. On the other hand, there are some issues, um, at, at-- that I'm more ambivalent about, right? So we have leading researchers in AI saying, "Look, we're, we're at...
[25:12] Richard Gold:
The frontier of AI is now dangerous, not just to states, but to humanity, and we need to slow down." But just to say, you know, we're, we're not gonna allow research in certain areas, I think, um, will undermine the flexibility of the innovation system. We never predict well where science is going to go, where technology's going to go. The history, you know, even the head of IBM doubted when they developed a personal computer that it would change much. Nobody, you know, the people talking about television when it first came out said, "Well, no one's gonna sit down on a couch and watch this thing except for rare occasions. They're gonna be moving about, so radio is going to, to live." Basically, all of our predictions about what humans will do with the technology have been wrong.
[26:06] Richard Gold:
That doesn't mean we shouldn't have systems to monitor it, that we should not anticipate possible negative outcomes and have a regulatory regime that would be alerted if these certain lines are crossed and take action. So it's not a, an excuse for inaction, but just to close down research in the name of research sovereignty, to me seems the wrong approach.
[26:29] Jonny Coates:
And I mean, that kind of leads us into talking about, not explicitly the situation in the US, but somewhat. So we've seen a lot of datasets disappear, public repositories, government websites, and so on. Regardless of what the reason for that disappearance may be, could you talk about any potential protection that CC licenses offer against losing those kind of things?
[26:51] Monica Granados:
Yeah. Yeah, it's a little bit sort of secondary. So a benefit of having any type of work be CC licensed is that the licenses are, uh, irrevocable. So that means if I create a dataset and I put a CC license on it, and then someone downloads that CC licensed dataset, I can never take the permissions that I have granted to that reuser away. So in this very, in this specific case of sort of datasets that are disappearing or, you know, are let's just say are no longer publicly accessible, if that original dataset has a CC license on it, it really gives you the permission to be able to have, uh, copies of those datasets in other repositories, because the license gives you that explicit permission, and that permission, again, is irrevocable.
[27:55] Jonny Coates:
So while CC licenses are great at protecting datasets and combating some of the geopolitics, what about AI and the impact of large-scale data mining? This has become a very contentious topic recently.
[28:06] Richard Gold:
Well, my whole research history shows that I believe openness is the way to go, and that closed knowledge is harmful. So I don't think it ever made sense, uh, to have too big a distinction. I think it, you know, it might go along with the stage of development, right? So closed knowledge about a particular application of, in, in a particular technology to provide exclusivity so that one has exclusivity, I can understand, because it doesn't really foreclose someone else from developing a competing product or better product, et cetera. Uh, but the, this fundamental... And I don't know if we've ever had a truly completely closed system. I think it's idealized as well. Lots of knowledge flows, uh, people talk, e-e-et cetera. But I think with the needs for large datasets to train AI and to validate the AI, it becomes harder to justify closed systems, because closed systems means idiosyncratic systems collected differently.
[29:12] Richard Gold:
It might be a high-quality dataset, but it's small, and to make it valuable, it needs to work with other datasets, and it's more or less impossible to do that if we have a bunch of different silos of closed knowledge. We need to open it up. So I think AI in general, uh, not just generative AI, but all the various forms of AI, are pushing us for, to large datasets, which naturally has an open aspect to it. And then the debate moves further down. At what point do we stop sharing? There's a, a strong argument for open source software, uh, behind the AI. What about the models and the weights and all the other aspects that go into an AI? And so I think that's where the debate is. Where's the dividing line? What is a particular application of knowledge that is best advanced through some form of exclusivity, and what needs to be open to keep the whole industry and research enterprise going? I don't think we have an answer on that yet.
[30:19] Jonny Coates:
Maybe could we talk a little bit from the stepping back- Place to begin with, and talk about how these have been trained on, on, um, a lot of open data, a lot of scientific literature in particular, and, and sort of what Creative Commons thinks about that. Um, is that a thing you feel was-- Do you think it changes what we think and mean by open going forward, or do you think actually things are all working as, as they were expected, this is just maybe a slightly cheeky use?
[30:51] Monica Granados:
Yeah. That's a great question, and a question that is really central to a lot of the conversations we're having at Creative Commons. I think it's incumbent on me to recognize what an impact the adoption and rise and sort of availability of large language models, commercially available models has had on the work that Creative Commons is doing. So CC licenses are copyright licenses, and in many jurisdictions and for many applications, large language models fall outside of copyright. And what that means is that it really puts into question whether or not the licenses are invoked in the use of, of CC license material for training. And so from CC's perspective, you know, the specific question around whether the CC licenses imply that their content is, you know, that there's some permission that is being given for training, from CC's perspective, it's that the licenses do not necessarily imply that the content is intended by the licensor to be used for AI training.
[32:12] Monica Granados:
It was, you know, for humans to use, but for new technologies to use it as, as well. And so I think what's changing for, for CC, and to get to your question about like, about open, is that I think in the context of AI, the question about open is now really fragmented by sector because of the, the scale at which AI is ingesting material, and ultimately about the lack of reciprocity in the system. So it wasn't a problem-- And, you know, these are, these are generalizations, but creators like artists or photographers who were putting things online, but their work wasn't being ingested at scale, it was only sort of being re-shared by humans. There was a, there was not this like outsized effect of their, of, of a potential, um, you know, jeopardizing, like je- jeopardizing their, their business. And getting nothing in return has a pretty big consequence for creators who rely on these types of, you know, on, on the, on their work to make a living.
[33:29] Monica Granados:
And on, you know, the other side of the spectrum, like researchers have a mandate to continue to share their work. They're adopting a lot of, um, AI workflows, adopting agentic, uh, or agents in their, in their workflows, so this is just sort of a part of how they do research now. I think there's a request for some reciprocity in the, in the form of, of attribution for researchers, but there's definitely just different relationships with AI from a body of creators who used to be more homogeneous when they were using the licenses, and now they're very distinct. And I think that does kind of change the question of open because now there's this sort of element of open may, may depend, or the definition of open might be different for different sectors.
[34:28] Jonny Coates:
So let's talk about some of those solutions. So s- signals is sort of the most visible thing I would say that you've, you've done around, directly around AI. So could you talk a little bit about what the signals are, what they in- what they were intended for?
[34:41] Monica Granados:
Yeah. It really is, uh, you know, a response to this lack of reciprocity in a system. The, the-- and the recognition that the licenses m- may not be fit for purpose it, for purpose for AI use, so for, for, for machine reuse. And so, you know, we started to think about, um, you know, first surfacing this like lack of reciprocity and, and really importantly, thinking about and challenging us and the community to think about what's the consequence to the commons if we don't intervene, both as CC and as, you know, uh, uh, the, the open community who is stewarding the commons. And so about a year ago, we produced a, a, a report that really spoke to this, the lack of reciprocity, what's gonna happen to the commons? Are people going to stop sharing? [upbeat music] We're gonna be issuing, um, updated guidance around how the licenses do or do not interact with AI and surfacing how that may be different in different jurisdictions, because different jurisdictions have specific provisions for AI.
[36:08] Monica Granados:
Certain jurisdictions have fair use or fair dealings and others do not. So this guidance, we hope, will really answer a lot of questions around how the CC licenses operate in an age of AI, because we are anecdotally starting to see that more researchers, authors ... are selecting more restrictive licenses in an effort to prevent AI from ingesting their material when all that action does is make it harder for humans to reuse their material. So that's our first intervention. Our second intervention is around making attribution the norm in AI. So, you know, Jonny, you, uh, very astutely pointed out that what researchers want is credit. They want attribution. That's the type of reciprocity that they want back for the, you know, AI using their work. They want credit for the, for the knowledge they created. So this work is thinking about how can we enable attribution in machine use or in AI systems, where is that being done?
[37:17] Jonny Coates:
Licenses are one way of providing some research security and in navigating AI. Monica mentioned earlier about governance and where the responsibility of CC ended, and as Richard alluded to, companies can decide to not be open with their data. But I want to broaden this, and so I asked Leslie about who is responsible for research security.
[37:36] Leslie McIntosh:
So part of it is each country's government is responsible for the security because they also have an economic obligation to their citizens to protect and to potentially, hopefully, not all countries do this, but to make their countries economically viable and sustainable for their people, particularly if you're talking about a democracy. I'm very much pro-democracy. So there is that aspect. I think also, and this is what part of forensic scientometrics is about, is I think there is a place for a group that's outside of a university or a company or a government or an institution, so like an NGO type, to both monitor and... I would just say monitor and help perhaps influence policy when it comes to that and when it comes to research security. So besides government, who have obviously economic interest in that, institutions have, have an obligation to this.
[38:46] Leslie McIntosh:
Right now, I feel like probably because it's the easiest to look at and it's the most exposed, one might think that publishers have a large responsibility in here. I am not saying that they are responsible free, but I think that the onus is other, is, is in another place. But as far as the responsibility, I'm gonna put it more at the institutional level rather than the individual, rather than the researcher itself. That being said, there are certain research fields that are very sensitive, that are more known to be sensitive, some that you wouldn't think of necessarily. The health field we've already talked about, and there's certain laws about that, but most of the laws have to do with protecting individual rights rather than a community. Other places that aren't biomedical, you talk to anyone who works with rare minerals, and they don't necessarily wanna share their data because then it can... they... those places might go and be, uh, destroyed, robbed. There's other risks to it.
[39:53] Leslie McIntosh:
Uh, anthropologists who study in particular individuals don't want their whereabouts known. So there's, there's lots of reasons that researchers might ha- want something to do with security, but I don't think it's security in the sense that most of the mandates are right now, which is about dual use and disclosing multiple affiliations and things, things like that.
[40:19] Katharina Miller:
So what practical steps would you recommend to a university now, today, to balance open science commitments with privacy and security obligations?
[40:32] María Bardají Cruz:
I think in sum, I would say, I will keep it very simple, be responsible. It's not that complex. I mean, it, it, of course it requires an effort, but if you consult the appropriate professionals and, you know, include that knowledge within your organization to help you understand how to use personal data in a responsible way, I think it would be easier for institutions to comply. So I would say certain measures, for example. So focus on purpose. What are you going to, to share the data for? Is it scientifically, is it, is it scientifically valuable? It is going to bring a value to society? Does it merit the access to the data? Keep control. Again, control. Use secure processing environments. Why not? Allow that access, but in a very particular setup with controlled environment, with secure environment. Think about, again, the life cycle, not just the access.
[41:35] Jonny Coates:
To, to round this out, looking ahead sort of five, 10 years' time, what do you think the relationship between openness, privacy, and security in science and academia will look like?
[41:45] Leslie McIntosh:
Oh, I wish I could say that it would be that idealistic picture where, where institutions have the resources for their researchers so that they could best share knowledge while adding trust to it. I'm not gonna say that's gonna happen in five years. I think five years people are gonna be complying with whatever security mandates are in place, and there's gonna be a lot of leakage all around. I do think, though, in 10 years that there might be less naiveté when it comes to the fact that the world was maybe as open as it was gonna get, and that we need to pull back and understand how this looks open and trusted simultaneously. And I don't think that you have to give up one without the other. I just think that we need to acknowledge it. And I know I've brought in a lot of things in here, but I also think about, like, an open river.
[42:44] Leslie McIntosh:
If you keep throwing junk into it and don't have any protections or regulations for not throwing junk in it, it's going to keep being polluted, and it's gonna be polluted even if you have those regulations. Then you have cleanup, and we need to do that when it comes to research to keep things open, available with understanding the acknowledgment that there is privacy, but privacy could also be used as an excuse for not having things open, and that the security of science, I'd like to think about that we move towards knowledge security and thinking about that from the academic space. But there will and are mandates coming through. The European Commission is working on a new one. Um, other countries are as well. That's not going away.
[43:35] Jonny Coates:
Open science and research security are often presented as opposing forces, but every conversation in this episode pointed somewhere else. The real challenge isn't deciding how much openness or security we want. It's deciding how we build trust, trust in data, trust in institutions, trust in collaborators, and ultimately trust in knowledge itself.
