This episode features audio from the Open Data panel at Open @ UNGA 81: A Conversation on Collaboration. It was recorded on September 21, 2026.
During July’s UN Global Dialogue on AI Governance, Secretary-General Guterres mentioned not letting “the digital divide harden into an AI divide.” This panel will explore the space between “open everything” and “close everything,” asking what meaningful access, governance, and participation require in practice. AI has changed the scale and stakes of data reuse. Material can now be ingested, combined, and commercialized far beyond its original context, prompting some communities and rightsholders to reconsider open access. But broad retrenchment carries its own costs. This includes a shrinking commons, disadvantaged researchers, and data access limited to those who can negotiate private deals. This panel explores the space between “open everything” and “close everything” and what meaningful access, governance, and participation require.
Announcer 0:02
Welcome to Engelberg Center Live, a collection of audio from events held by the Engelberg Center on Innovation, Law, and Policy at NYU Law. This episode features audio from the Open Data Panel at Open at UNGA 81, a conversation on collaboration. It was recorded on september 21, 2026.
Angela Oduor Lungati 0:25
Thank you, everybody. It's always fun having a panel right after lunch when people are jet lagged, right? My name is Angela Odolungati. I am the executive director of Ushahidi, which is a global tech and data nonprofit that's been turning what communities experience and witness into data that counts since 2008 in over 160 countries. I also have the immense privilege of working with Anna and team as a member of the Creative Commons Board, actually as chair, and really excited for today's Conversation. So I think in July, during the global dialog on AI governance in Geneva, the Secretary General warned against letting the digital divide harden into an AI divide, pointing to how little is being invested in AI capacity for the developing world. I think that warning was really about data, because we all know that AI systems learn from data that people and communities create, and often from one audience, one purpose, one place. And that material is now being ingested, combined, and commercialized at scale and at a distance its creators never imagined. And for 20 years, a case for open data has really rested on public benefit. This idea of transparency, research accountability, and the ability for many hands to build on this shared knowledge. And I've spent my career on that promise. But AI is changing the stakes of reuse, and so there's some communities, institutions, and right holders that are asking us very hard questions. What should we still make openly available and on what terms? The easy answers are open everything or close everything, and this panel is about the space in between. So, what meaningful access, governance, and participation actually requires, especially where data is culturally sensitive, underrepresented, or vulnerable? And I've got three amazing people on this panel who have each built a different piece of that answer. And I'll start off with Chris. I hope I don't butcher my introduction. I'll still give you space to talk a little bit about the work that you do. He's led Public Knowledge for the past seven years, and for 25 years, Public Knowledge has been a leading voice in DC for open tech, digital rights in comms, tech and media systems. Welcome, Chris. And then we have EM Lewis Jong, who's a founder and CEO of the Mozilla Data Collective, which is a data sharing platform that enables creators to share data on their own terms and determine what value comes back to them. So, looking forward to having a conversation with you, and we have Jane Anderson, who is an associate professor here at NYU, the director of the Indigenous Data Law Lab, and founder of the non-for-profit organization called Local Context. I mean, for 15 years, Local Context has been providing these Indigenous communities researchers and the institutions that are labeling this well, giving them data labeling strategies for Indigenous provenance. So, looking forward to this conversation as well. Now, the way this panel is going to flow is, we're going to have an initial conversation. I don't want us to be talking at you. So, at some point, I will open up the floor to questions, and then we will wrap up, or at least try to. Even though the topic seems very, very heavy, so I will start with you, Chris. For most of the open data era, access always meant you know a person could read and download or build on something. But now the primary consumer of content that's being published online isn't a person; it's a model. I think AI is even replacing such as you know how people are able to find and interact. What it is that we publish. So when everything is treated as training data, what assumption inside our open data policies break fast?
Chris Lewis 4:19
Thank you for having me, and Anna. Thanks for inviting me. I actually have four things I think, but they all I think circle back to a level of trust. The first one that we're following very closely is the breakdown in kind of the. It's not an actual agreement, but the way that the open internet has worked around copyright law, we're big defenders of public knowledge of fair use as creating benefits for democratizing knowledge and information on the open web, and that's being challenged right now because of the scope and the scale of AI. By folks in the in the publisher and content creator community, and and we think that could threaten the way certainly in the United States where we focus our work could threaten the balance of an open web if you see a rebalancing of fair use analysis and copyright law in the wrong direction. We think AI training should be a fair use, and that's being litigated in the courts right now. There's certainly a breach of trust around consumer protection and privacy, and cybersecurity and safety, and things have been brought up in other panels, I think there's someone mentioned the lack of systems we have for traffic being mostly bots instead of humans now, and we we there's trust inherent in that that when you move on the web or when someone queries your page, like the the Wikimedia problem, that that it's not extractive or overly exploitative, and then one last one that I think people are missing, or that we haven't really talked about much today, is the trust that you can use this technology and it applies to your life well. My background in tech policy actually started with broadband digital divide issues. I worked at the Federal Communications Commission here in the United States, and and we're concerned about what I think is starting to be described as an AI digital divide that will have some of the same principles that you saw in a broadband digital divide globally and in the United States, but it will look different in its application. But but you know questions of do you have access to the technology? Can you afford the technology as it becomes more and more expensive? Is there an adoption of the technology by folks who aren't in this room, folks who are marginalized, low income in remote areas, and and so all of the I think that goes to trust and people trusting that they can use technology and have access to it, but also that there's a public interest balance of of values in how it's used and applied in society.
Angela Oduor Lungati 7:22
June. move on. Move on to you. So your work starts somewhere else with provenance, consent, and community authority built into the data itself. Now, for a room that mostly thinks in open versus closed, what does data sovereignty actually mean, and how is this indigenous or community governed data different from the default open data mindset.
Jane Anderson 7:47
Again, thank you so much for having me here and to be part of this conversation. I guess where the work that we do from an indigenous communities standpoint is thinking a lot about how that data has already been extracted, and that Indigenous communities' names and attribution and provenance is not part of the data. So it becomes very difficult to exert governance over that data in terms of how it should be used, how it kind of meets different kinds of communities' expectations about how that data can be used, and then of course, that has all these flow-on consequences. So, a lot of the work that we've been doing over the last couple of years has been, you know, firstly, how do you build? I would just also say that copyright law does nothing for Indigenous communities when that information has already been extracted and is owned by other people, so you know we have a particular problem in relationship to the way in which copyright serves Indigenous interests to start with. So what we've been doing for the last couple of years is actually building different kinds of mechanisms that can get into the data infrastructures that actually connect Indigenous communities' names and protocols and provenance information back into the systems that have historically excluded that. That's quite complicated because you have to build new fields, or you have to argue that fields that have been built that never thought that these issues were something to think about have to kind of redesign themselves. But what we find is that that key provenance piece, which we know is a big problem for AI itself, actually has a predate problem, which is also how Indigenous people have experienced the extraction of their data. So, if you kind of run these lines of I'm going to say the word colonialism through particular kinds of spaces, we see that that that deliberate erasure of provenance has an antecedent that serves a certain kind of accumulation of data in another space, and so the tools that we've been working on include. I mean, I would say we call them kind of extra legal tools, partly. Because we see that extra legal tools support the development of legal tools down the line, potentially, not the other way around, and so you know that's why we have built local context as particular kinds of tools that can allow for communities to assert their interests back into data systems, but also work in the standard space as well. And we just created the first standard in this area, which is actually around provenance, provenance of Indigenous peoples' data through the IEEE, which did take us five years to create, because we're not seeing these interests that come from already marginalized communities actually ever really being elevated to a space that they're taken seriously, even though some of the the the reasons why that erasure has happened in the first place are actually very relevant to our conversations today. So I'll just stop there.
Angela Oduor Lungati 10:58
I have a follow up. So the start. and the work that you've been doing is essentially holding AI systems and humans-you know, people who are data actors-you're holding them accountable for where, you know, for recording where the data come from. Comes from. So, based on the work you've been doing over the last couple of years, what becomes possible once provenance is a requirement, not an afterthought.
Jane Anderson 11:26
Relationships become differently possible, and I'll just give you one example. Maybe it's it's kind of a good example to speak through. So one of the biggest biodiversity kind of holders in the world, this is the Global Biodiversity Information Facility holds, I think, over 5 billion occurrence records that has been accumulated from about the 16th century to today. And if you kind of look at a map of the occurrence records, kind of filling the globe, it does look, you know, quite striking in terms of you know where where there isn't an occurrence record is really the question now, and when we try to look at that data to find indigenous data, because finding indigenous data is kind of key to indigenous data governance. If you're familiar with the indigenous data sovereignty spaces, governance of data is kind of the precursor to doing anything else with data. It kind of starts from data governance, but you need data in order to exert governance over that data. And if you don't know where that data is, which is largely the case for indigenous collections and indigenous data, it's very hard to assert governance over it. So here we are. We're looking at this global biodiversity information facility to try and identify where is the indigenous data in those 3 billion occurrence records, and there was no way of finding it. We deployed all sorts of tools to try and find where that was, whose it was, and we couldn't come up with anything because partly the metadata fields themselves didn't hold that information. So, over the last two years, we've been working with GBF to kind of change their metadata structures so that they can include different ways in which communities add their provenance information, sometimes using local context labels. And now we can see 2 million records that come from Indigenous communities. Now that doesn't sound much in relationship to 3,002,000,000, but what it does do is it elevates communities' names so that they can be part of a conversation where previously they were never part of a conversation, and it's that that enables a different kind of possibility of relations and conversations of communities that are just largely just kept out of those spaces, and so these are small initiatives, and they take a lot of time, as I said. But they are really about building. Well, why does provenance matter, and how does provenance matter, and to whom does provenance matter? And I think that's kind of where local context really sits within this space, particularly if you don't have a legal framework that you can rely on, because that was never part of the framing of your interest to start with.
Angela Oduor Lungati 14:13
So I I do want to move us on to the tools or what the tools actually do. But before we go on that thread, I have a follow-on question for you, Chris. So, I mean, sovereignty gets claimed at every level. You know, we're just talking about it from a community level, indigenous nations, institutions, and countries. I'm just wondering whether this scales up cleanly, or does you know national data sovereignty claims sometimes cut against the community's authority,
Chris Lewis 14:44
I don't think it has to.
Speaker 1 14:45
Okay, it
Chris Lewis 14:46
may right now. If you look at where the power lies, you know, unlike the early internet, AI is being developed largely in the private sector by large companies that have a profit motive early. Internet had open protocols because it was not developed in that way, and then you layered on business models and profit onto those protocols. We're just not seeing that this time, and so there's great power. I mean, just in the narrative alone, we heard on the last panel, there's great power, you know, that rests with the companies that are developing the technology that have the capital to hire the best experts, we've heard this throughout the day. Then you have the geopolitical power. You know, it's why you know it's it's why I don't use the term data sovereignty in the United States. To be frank, as an American who only really works on U.S. policy, and given the the atmosphere in Washington D.C. right now, it sounds like a geopolitical race, and that's not what we need as advocates for open technology. We need to make the arguments for the the proper structures for flow of data that shares, but is responsible, and and so we're trying to make a case for openness in a very you know geopolitically difficult town right now that that doesn't sound rivalrous with middle powers in Europe and other in other places. That's a challenge. So we talk about you know what are the benefits of openness to a a multicultural society that's large like the United States, and that's where you could use the term sovereignty. I just don't. I think it's interchangeable with empowering local communities to have systems that allow them to yes get provenance and credit, and to have a bargain that works for them with open access to knowledge and information. So you could use the term. I don't, but I think it flows. It it has to work at all levels of both geopolitical politics all the way down to the local level. Yeah,
Angela Oduor Lungati 16:59
and I appreciate your caller for it doesn't have to be so. Just because that's what the status quo is, that that doesn't you know stop us from really reimagining what it should look like. Thank you for that. Yeah, moving on to what the tools actually do. I think there's a temptation to think that a smarter license will fix everything, and I know that you all have built an actual space where communities set terms, you know, keep ownership and determine what value actually goes back to them monetarily or otherwise. So, from the inside, what can licensing genuinely accomplish, and where do you see, or rather, where have you seen it hit its limits for it needing something else, whether that's technical controls, provenance systems, data trusts, anything. Yeah.
E.M. Lewis-Jong 17:45
Well, I heard a chuckle go around the room at what it was really loud. It was so loud. It was like, yeah. Let's all get our licensing cards out. That'll be a fun conversation. So, licensing is a tool in a toolkit. So we sometimes talk about this as you really need a stack of interventions to guard utility of data, and anyone who tells you there is some magical technical solution that will make governance super easy like this is full of it. It's not true. Like you need legal safeguards. Licensing is one type of legal approach, one of many. You also need technology that is going to give you a licensed teeth. There's no point putting a license up if you don't have authentication systems to see who is using the license. We see often on. I'm not going to name particular names because I so incredibly don't want to go to prison. But one of the one of the impacts we've seen recently is actually some of the most shameless actors will be the ones who violate license terms, and they'll see that a community has stipulated X license, and then they'll go ahead and use that and play catch me if you can with us, and then actually some of the really conscientious and thoughtful AI model builders will not use the dataset at all because they're worried about whether or not it contravenes the community intention. So that creates a very perverse incentive system. So you need to be able to pair licenses and other legal safeguards with technology behind the scenes that allows you to identify actors who may need to have sanctions to prevent them from accessing the datasets. We encourage anyone who has particularly, shall I say, artisanal desires expressed in their licenses to actually put it behind a gating mechanism. Don't just stick a license on the datasheet; it doesn't work. You need to have some type of system or flow whereby you release the dataset once you are satisfied with the request that's being made, and then the infrastructure to automate that over time is exactly the type of thing that we're doing at the moment because it works really well if you're a small organization to be sitting there manually approving requests. But if you're a steward of 100-year archive, you can have 1000s of data requests in a. And so you need to have systems that make that type of governance easier and more scalable. And then the third piece is community, and I, I'm surprised how little we talk about this because actually, like collective bargaining and community mobilization and us being transparent with each other about things like license prices, those things are enormously powerful. They're not legal things. They're not technical things. They're working together things. And I think one of the pieces that we don't talk about is actually AI companies really need us. Like they really need data, and they can go out there and be like, "This is my pricing sheet all day long. But if everyone stopped selling to them, they would be screwed. the the The quantity of human content on the internet is declining over time. Their first approach to training models is not sustainable, and now they are going to extraordinary efforts to get datasets from other sources. And they play hardball, like they don't need us, but actually they need us quite a lot, arguably more than we need them. So I think, like, it's so incredibly in their interest for us to all feel disempowered. We're actually not disempowered. Like, data stewards hold an incredible amount of power, and it's really about leveraging it.
Angela Oduor Lungati 21:13
Before I pull on that thread around how the AI companies need us more, first of all, powered people, I want to. So I know you've talked about the bad actors trying to cut corners and everything, but if you look to what an ideal future looks like, what is it that you think actually changes behavior? Will change behavior versus what it is that looks good on paper but nobody honors downstream?
E.M. Lewis-Jong 21:37
Yeah, it's such a good question. So I think, firstly, from our perspective, biodiversity. I feel like you've really started a thing, like rewilding biodiversity, more people training models that are right size for their use case. I think is a much healthier ecosystem, and so to the extent that data stewards can help to seed that, I actually think that's a that's a really important like piece that we get to do here, in terms of stuff that looks good but maybe doesn't matter a lot. I think sometimes it's just about the maturity of the things we put in place. Like the EU AI Act has now stipulated clearly that people need to disclose their training data sources, and I feel like I'm amongst friends. So if anyone has read the disclosures, they're extremely funny. Some people, it's actually, I mean, it's masterful the extent to which they can say nothing over a very long period of text. It's like, it's like, wait, where is the disclosure bit? So I think it's a, you know, it's a new system. This is the first round of disclosures that's being made. I would expect the next round to be responsive to policymakers saying, actually, these are all the loopholes that people tore right through. How can we make the next round of disclosures better? And even though it's very frustrating for those of us who are in it to see that this process takes a bit of time, it does take a bit of time. Some of these measures are very new, and I think it's understandable that the first round are maybe not as not as powerful as we might have been hoping.
Chris Lewis 23:00
We haven't heard the the concerning side of licensing here, and I think it's important to note that it's hard to calculate, but it it does not seem like there. If you just look at the trillions of dollars in lawsuits against AI companies for for you know claims against use of content the the that there's enough money to go around that this somehow changes and and and actually replaces the the losses of revenue in the market to creators and others. I just don't see the math working out that way. It's not to say you can't do licensing, but but there's there's an expectation that there's this pot of gold somewhere that I just don't think is true for the average creator or or publisher. I think it's especially true for small ones that don't have the leverage to negotiate like the New York Times does and other large players. I would also add that you know there's a backlash happening because of of folks feeling exploited that is weakening the democratic access to knowledge information online. People are locking down every way they can. They're trying to lock down data and information on the open web, and the biggest AI companies are cutting deals to have access to to data from specific sources, and it could really end up with a two-tiered you know, digital divide, where if you can afford to pay for the services of quad or a major company, then you'll have access to the data that they pay for at a price. And if you can't, you're going to get AI slop. You're not going to have the open web because people are locking down access to the open web in the reaction. And so we have to find. A new balance that creates structures that continues to leave knowledge and information democratized, and so you know these are the these are the threats of licensing licensing right now, and and we have to strike a balance there.
Angela Oduor Lungati 25:16
So you've actually answered the next question. Oh, sorry. No, this is good. This is good. I mean, it's it's a natural reaction to kind of feel like you know you're exploiting everything. I'm going I'm going to close up. And EMU said that the AI companies need us more than they have more to lose, but I'm I'm going to frame that as a challenge. And this is to everyone except Chris, because you already answered that question. Who exactly is losing here? Is it is it just the dominant air companies who can, to his point, write a check and get a private deal, or is it the research startups, the public interest technologists, and the local developers who can't? Who's the one who's really losing here?
E.M. Lewis-Jong 25:55
Yeah. Well, I just want to I want to make something clear that I I think is very clear, but maybe is not clear. Which is, you don't have to give one license to everyone. That's not a thing. Like you can have one license that you execute for your relationships with OpenAI, and it does not need to be the same as the one that you have with your local community. And so, one of the things that we set up very deliberately to do with MDC was give people the option, and most people actually want to have gradiented sharing, where they're more than happy to make their information publicly available to groups who they think have their best interests at heart and who are going to steward that information with care. They don't really see why they should subsidize better's AI sourcing strategy for free, and I think that's a fair question. So I think completely agree. Like the loss of the open internet is something that all of us should be thoughtful about. But let's just be really clear: who created that problem? It's not creators, and I like. I think there's a version of this that ends up sounding like victim blaming.
Chris Lewis 26:54
Oh no! Absolutely. No, I mean these are real reactions happening in the marketplace. People, you described how people are looking for tools to lock down their content. I'm not saying that I'm blaming the victim. This is what's happening in the marketplace. Yeah, I think we have to find structures that strike a balance that we don't see that closure and closure. That's what I'm saying. I don't think it's blaming the victim at. I think I'm describing the market.
Jane Anderson 27:21
I guess my like the the different kinds of like spaces that I work in is is a little different, where it's not not kind of necessarily locking down data that you can you already control. It's finding the data so that you can can create different kinds of possibilities for governing that data. So, if we were to find another word for other words for sovereignty, I would say kind of governance and decision making are kind of key axes of what sovereignty is about. And for indigenous communities, one of the and you know I apologize in these contexts. I am a settler. I work for Indigenous communities across the world, thinking about these questions of data sovereignty and who owns Indigenous data and kind of what does that mean. But it is creating the possibility of different kinds of trust and trust relationships. So we do a lot of work building Indigenous data trusts, which is bringing data that is completely disaggregated in multiple different kinds of places together back into one place, so different kinds of local governance can be determining how that data is used for different kinds of predictive effects into the future, and so again, it does work at a kind of more local level in that sense, but it actually meets some of these kind of key community-based questions in relationship to what does the health of our waters look like? How do we get all the data that is being accumulated about the kind of our waters and homelands over this long period of time back into a context that we can make decisions about and have different kinds of questions about what is going to happen within these spaces for future related climate issues that we're not going to get from a government agency. So there are kind of different ways of thinking about what does that governance or what is the data trust and that trust piece is really important. That does actually require both different kinds of technical infrastructure and different kinds of legal infrastructure that allows for communities' expressed expectations around that using data to be differently realized. I would say in the other space that I see indigenous data being highly valuable is indigenous languages obviously, and all the work that is happening to both secure and hold and you know reuse that data. It's because often that indigenous language data is not sitting within indigenous community contexts. It's sitting within. Museums, archives, libraries, universities all around the world, and so the kind of rush on indigenous languages is extreme, and yet there's no governance happening around that. So that is what pushes people to be thinking about how do we create our own indigenous data trust? What does that look like? Because we're looking about looking at kind of return for access, and then deliberative decisions in relationship to that, and that gets to these different kinds of questions about what does deliberative democracy look like within the data spaces. I'm going to ask a slightly provocative question. So there's no denying there is, you know the the idea of a shrinking commons is very very real, and I also wonder whether that that term shrinking the commons is the right frame when you're dealing with communities that were never fairly represented in the first place, and where you know the risk isn't necessarily losing access, but it's facilitating
Angela Oduor Lungati 31:00
extraction, and anyway, it's a question to everyone. But I'm specifically asking you because of the work you're doing around protecting indigenous communities. How how do you deal with that? Like holding those two treats in one. It's
Jane Anderson 31:13
a great question, and you know, our work has never. There isn't a fuzzy, warm glow that goes with the commons, because the commons has been a site of extraction. Indigenous people have not historically been the public of the commons either, and so there is a certain kind of disjuncture that sits in relationship to upholding the commons when that commons has been a site of extraction, and you know it doesn't get talked about very often actually. And so I really appreciate your question because we're just trying to claw back space into what is the commons. The Commons is almost a cultural construct that has historically excluded certain people over others, and we don't talk about it as a space that has embedded colonial histories and powers as part of its very makeup. And I think we do need to talk about that because that then means that we can look at those who have been marginalized, or indeed their data is part of the commons, but they have had no say in how that data came to be put in that space to start with. If it isn't even a space,
Chris Lewis 32:30
can I just take that and turn it sideways? Because I would argue, and and and you know more about this than I do, but tell me if this sounds right. I would argue that the Commons from 1995 to now better than the Commons before it, because the interactive nature of the internet at least allowed community building and more democratized access to a Commons that is still, like you said, controlled by folks with power, libraries, you know, folks with capital. Where before it lived in regional silos. What I'm saying is, how do we build structures that actually shares power with local community, and so you you continue to improve the power dynamics that did improve with the internet, but we're not good enough. You know, to me, local computing power, local models that are done in collaboration with local communities allows people the power to say what they need from a model, what the the data practices are, and then and then share that with the world on fair terms, but also preserves that local culture, history, language. Hopefully, I'm saying local, and I think geographical view, but it is really about community that doesn't have to be geographic or yeah. So, I think when we talk about what do we need to build in that direction?
Jane Anderson 34:02
Yeah, yeah, yeah.
Angela Oduor Lungati 34:03
And I'm appreciating you bringing that thread so that we're not just focusing on what the challenge is, but then what the opportunity also is, which actually leads me very well into the next conversation. Ian, you talked about having different licenses that you don't necessarily have to have the same one. So, you know, can we actually build differentiated access? You know, a model that keeps that public interest and research open while giving the data stewards real control over high commercial reuse, or is that just a nice idea that collapses the moment the data is out? And specifically from your experience with MDC,
E.M. Lewis-Jong 34:39
yeah, it's like as with so many things, prevention is so much easier than the cure. I think to your point about pots of gold and after the fact, I agree. Like, I am not advising anyone who uses MDC necessarily to start hiring lawyers. I think that ship has thoroughly sailed. Buy ship. The now the question is, what do you do with the 99% of. Assets which are not on the public web, like because we should just be really clear, data abundance is not going to come from the open web, and I think nothing that we probably say in this room is going to make people feel more safe to do that at this particular moment in time. I live for the day when people find the internet to be that space again, but I don't think it is right now for a lot of communities. So the question is, can you make it possible for people to share in gradiented ways? Yeah, absolutely. I think actually we already have a lot of the technical infrastructure and a lot of the legal infrastructure and a lot of the community infrastructure, and it really becomes how do you stitch it together so that it actually works kind of cohesively. So, for example, we're talking a lot about licenses. Licenses is just one example. Trusts are another one. Cooperatives are others. There are lots of vehicles you can use to set and state intention. This is what I want to happen with my data. There are lots of different ways to say that. Licenses is just one. Okay. Once you've done that, how do you build out technical infrastructure to make it much easier for you to make sure that when you're handing this data out, you're sure about the person on the other end of the request? And again, we have a lot of the tools to do that, and it's just about bringing them into platforms that make it easier. And then the third piece: the okay, what happens in terms of community action when something does go wrong, because things will always go wrong, and I think that is really about mobilization and that layer, that kind of mobilization and community sharing layer. I think it is one of the things I see as a potential, not drawback, but maybe risk in some of the execution of sovereignty. Is that I also see a lot of silos, so you have a lot of different people, all maybe striking deals or building infrastructure. None of them talk to each other, which means that they and the only person talking to all of them is the downloaders, and so that really allows them to set the terms because then say they say this is what the price is for blah. We've never paid more than that. It's always that. It's 5p, and and like and how are they to say that that's not true? So, you know, I think galleries, libraries, archives, and museums are one example of this. They're cultural stewards, and they have these mandates that sometimes pull in different directions. They have a mandate for public access, but they also have a mandate to steward the collections. And stewardship and open sharing are not always, at the moment, easy bedfellows. How do you make it possible for organizations like that, who are trying to strike sharing agreements or build out tools of their own that might have their own sustainability models? How do you make sure that they're all talking to each other, such that information is is flowing freely? Because I really think that is part of the answer, and maybe where we would find kind of a lot of common ground is like I'm really curious about could we have more transparency indices out there where people actually disclose the types of arrangements that they're entering into, or where people disclose these are the types of terms, these are the types of price points, so that everyone can actually see what's happening in a much more kind of open forum.
Chris Lewis 38:02
It's it's like Michael Weinberg was talking about with the open hardware that it works because it's a community. Yeah, and you're talking about community norms, but that need structures to facilitate that community. Trust. Yeah, this is why. you know I I understand, but shy away from the term of digital sovereignty because it it breaks down the sense of a global community, and you know where is the effort you know you know it's it's it's it's ungo weak here in New York. Where is the effort from global institutions to set up trust-building structures that allow that data to be shared and still respected.
Angela Oduor Lungati 38:48
We could go on and on, but I really didn't want to be the the group of people just talking at you all. So I want to open up the floor to maybe one or two questions before I hand it back to the panelists for final remarks. Anyone? Yes. Please come straight to
Speaker 2 39:04
the.
Angela Oduor Lungati 39:05
Okay.
Speaker 2 39:05
Okay. So this one is particularly for Jane. Thinking about your your work with indigenous communities, I think it's great that there's consideration for the ways that their data is being used, and also focusing on their language preservation through that, which of course is incredibly important. But thinking about the climate implications of AI and how important land and natural resources are to indigenous communities, what ways Are you or just in general? Are we working on finding a way to bring those things together so that the you know the the ways that we're accessing their data and using their data isn't also further destroying their land that we took away from them.
Jane Anderson 40:00
Yeah, thank you so much for that question, and I'm just going to say we do have a panel tomorrow night called "Living Data, Living Lands" that is kind of thinking about some of those questions, and particularly across across the world, those are the kind of key questions. We have a kind of a framework that we kind of call relational infrastructure, and a lot of our colleagues in Atero and New Zealand have some very kind of good examples of how they actually do that in different kinds of ways. That are, I guess, and thinking about it being less extractive of the data and of the resources that are required to be stewarded for the future. And I think that there is this interesting question about where does the future generation sit within some of these conversations, and what gets built has consequences. And so, you know, these are very small scale projects, but also what we're seeing is different kinds of tribal nations coming together, where one tribe might already have a data center that allows other tribes to utilize that data center rather than having to build their own one on their own tribal land. So we've got these really interesting cross-jurisdictional land questions that are coming up, and I mean, so much of the work we're doing is also educating tribal government around this this whole space, because those decisions need to be sorry natively sovereign within those spaces, and that requires again different kinds of information around how those decisions can get made, and who benefits from them. So the land piece is a huge piece because we see that all these pieces are related to each other, and they have long histories of overlapping in relationship to what can be done and who gets to control the kind of outcomes of them. Come to our panel next week. No, tomorrow night.
Speaker 3 42:01
Jane
Jane Anderson 42:02
is
Chris Lewis 42:03
obviously the expert here, but tangentially related in the broadband space in the United States, we have worked with tribal communities to not only get a set aside of spectrum. You know, this is a national public resource, but tribal communities were not getting broadband. It's the least built-out broadband place, so where the digital divide is the greatest in tribal communities, and so we got tribal priority windows before Spectrum was auctioned off to the big companies, and at the same time, there are communities, Institute for Local Self Reliance, and others who are helping train tribal communities to build infrastructure in the way they want, so that they own the spectrum, they own the resources, and they build how they want in their community. It's the same concept, just how do we apply it in the AI age? Because there are right ways and wrong ways to build data centers.
Speaker 2 42:55
Before,
Angela Oduor Lungati 42:59
so let me just check and see if there's anyone else in the room who has another one, and then Clay. Do you mind using the mic? Thank you.
Speaker 4 43:15
All of you serve offer a service to your communities. All of you also are have talked about trying to get communal involvement from the people you work with. Is there a tension between the way you think of your organizations and the way the people you serve think of your organizations? And if so, how are you addressing that? Thank you.
E.M. Lewis-Jong 43:38
I can go ahead because I'm so keen to answer it. Well, so I think Mozilla Data Collective is self-service, so people only join if they want to, and if they don't want to, then they don't. So I think that that solves a lot of problems. It's the reason that we don't allow scraped content, we don't abide by fair use. It like if you don't own the copyright and you have no legal claim on the data, you don't put it on the platform. We won't allow it, and so I think in that sense, it's very much like here is a platform in the classic sense. It's yours to stand on. I would say sometimes I think on these panels I come off as like anti AI builders. We're actually not like we are trying to solve the data problem for people who build technology, and many of them think of us correctly as like friendly because we do want to help them solve that problem with consentful, diverse data rather than with data that they procured without consent and that looks like Reddit. Like we we think there's a better way to solve this problem. The vibe is more like let us help you than antagonistic actually.
Chris Lewis 44:45
I think for an org like public knowledge, who has fought for open internet principles and you know especially fairies for 25 years, the challenge is when someone is feeling threatened. And we often offer a nuanced policy solution that says, "Well, you need both privacy, tech accountability, and openness through fair use. It's too much. Someone feels like they're under threat, and they're like, "Why are you arguing for something like fair use that big tech agrees with? They're arguing for it because what big tech created, or really what big tech exploited as Web 2.0 off of Web 1.0, was the open protocols and made money off of it. What we're trying to do is say, let's you know, let's find a way to keep the benefits of openness while having accountability, and then I think that's often too nuanced for people who feel under threat, I think it's intellectually correct, and so that's the challenge that we face. But I believe in it.
Jane Anderson 45:49
Thank you for your question. And I think we, at local contexts, we serve three different kinds of communities. We have indigenous communities, we have institutions, and we have researchers that are working with Indigenous communities, and you know, from our perspective, we don't have enough capacity to respond. And you know, but everybody has different like needs, and most of our work goes towards supporting communities first, Indigenous communities first. But institutions also know that they're they're sitting on a certain kind of precarity of their collections, particularly colonial collections, that they can't really share because they don't really have the proper provenance. That means that they can do either meeting the kind of care principles or even developing the trust they need to have with the communities that are going to use that material. So I would say we kind of our capacity is kind of always at overreach because the need is so great, and that is a need that's an international need, not just a US-based need as well. So yeah. Were you looking for an answer from me today?
Angela Oduor Lungati 47:00
Well, yeah, I know, I know. You know what? I'll answer you personally after this, just because I really want to give the panel space to like wrap up. I know we've talked a lot about all the challenges and all the problems, but if you were to think about, you know, picture a healthy commons in the next five years. Name one thing or one choice that we have to make now to help us get there. One each that could be policy, investment, governance, or infrastructure. Just one thing we need to invest in now to give us a healthy commons in five years.
Chris Lewis 47:36
I work in the policy space, so I'll try. Okay. The policy front, I would love it if the dysfunction in Washington allowed the big one of the biggest players in the country in the world to invest in and reimagine what public media is, and to include AI infrastructure as part of that definition. We saw it in the 50s and 60s when you know we created parallel public infrastructure for radio and television, and and that investment went away in the last couple years, and so you no longer have a public serving infrastructure for radio and television, or or it's it's under serious threat because one of the funding legs went out from under the community. But if we did that with AI, it could be with standards that hopefully the community would buy into, and I mean all communities. Thank
E.M. Lewis-Jong 48:30
you, Ian. I'm not a policy person. I, I think if I could wave a magical wand right now, I would get open source AI and public interest AI to have a go to market strategy because it's not the case that the models being produced by the corporate labs are any better at most of the tasks that we use them for every day. And there are, as we've already heard, there are lots and lots of really thoughtfully curated, well trained public interest and open models out there, and like not that many people use them, and not that many people are thinking about the things that we're thinking about. I think this needs to get into the public consciousness as a valid alternative to getting an account with ChatGPT, and it's just not it's not there yet. So that's the thing that I would love to see. Is like there's incredible work being done. What is the wrapper that makes it consumer friendly, and what is the what are the resources we can mobilize to get it into the hands of more people as a viable alternative to enterprise AI?
Jane Anderson 49:33
And mine is a little bit infrastructure. I think a lot about that. I want to see infrastructure that really allows for proper, transparent provenance, but also for the capacity to signal preferences in how different how data is going to be used, and that kind of preference piece or that signaling is really key to kind of knowing down the line how that you're using it ethically, responsibly, and with care. Okay.
Angela Oduor Lungati 50:00
So, wrapping up. Thank you, Anna. I see you tweeting your thumbs. I had a few things. I had, you know, one access is not capacity. You need awareness and education. We need stewardship. We need participation, and we need agency. Cool. Thank you all very much. Please don't kill me, Anna.
Announcer 50:26
Engelberg Center Live Podcast is a production of the Engelberg Center on Innovation, Law, and Policy at NYU Law, and is released under a Creative Commons Attribution 4.0 International License. Our theme music is by Jessica Batke, and is licensed under a Creative Commons Attribution 4.0 International License.