Welcome to CDT's Tech Talks, where we dish on tech and Internet policy while also explaining what these policies mean to our daily lives. I'm Jamal Magby, and it's time to talk tech. Welcome to Tech Talk. Bye.

Today, we're thrilled to be joined by two voices at the forefront of digital rights and open technology. Mark Sermon, president of the Mozilla Foundation, and Alex Givens, president and CEO of the Center for Democracy and Technology. Together, they spent decades advocating for an Internet that is open, accountable, and centered on human values. In this conversation, we'll dive into some of the biggest challenges facing the digital world today. From AI governance and privacy to platform accountability and the role of civil society, we'll also explore what it will take to build a web that truly serves everyone.

Well, super happy to be here, and I've always loved working with CDT over all the different eras of of Mozilla. And I like that you characterize us as a Marvel Universe. It is a little bit like that, in that Mozilla is a growing cast of superheroes. And so, you know, we still have the Mozilla Foundation at the heart of it, which was the organization that started, you know, well over twenty years ago now to make Firefox or really to make sure that the web stayed in public hands, of which Firefox was the first big thing we did. But now we are a set of companies, venture funds, other nonprofits, all aimed at different parts of how I think about it is doing in this AI era the same thing we did in the the web era, which is build things that push the trajectory of tech, the trajectory of our relationship between tech and society and people in a better direction than they would go otherwise. And I think they're not going in a great direction. So I can go through the the different pieces, but, you know, you have consumer products like Firefox and Thunderbird evolving into how can they play a role in shifting AI in a a better direction. You have more developer oriented things, a new company, Mozilla AI, a new data oriented company. We have a venture fund. Of course, we have all of the different fellowships and programs that the foundation itself does philanthropically.

And Mozilla was just such a amazing project. And it really, in the beginning, wasn't even an organization. It was this project built around, in part, the the Netscape code that had been released as open source. But it was a bunch of people who were trying to make sure that the whole stack of technology, you know, JavaScript engines and bug trackers as well as a suite of browsers and emails grew up to compete with Microsoft, which had started to dominate where the web was going. And so, you know, I was a fan of, like, look at this bold effort of people who are trying to take on the biggest tech company in the world and send things in a different direction. And, so I guess, I mean, in many ways, the proudest of the fact that Mozilla even began from that, taking a different approach, a community based approach to taking on, you know, such a a big kind of David and Goliath fight. And and over the years, building on the success of Firefox and and the success of Thunderbird as a open source privacy protecting email client, we've really tried to then figure out what's next. And I would say there was a period, I came, you know, just before 2010, where we were trying to figure out what's next from a tech and a product perspective. We spent a lot of years trying to build an open source smartphone that we compete with Android and, and iPhones. Didn't work. Didn't work for a lot of people who tried to do that. But the other bet, and this is what I was doing as executive director, was to really grow the Mozilla community, not just to be the nerds or to be a lot of different kinds of nerds, and to bring in journalists or educators or scientists who had this open source ethos and thought we have a role to play in taking the digital world and the relationship between technology and society in a different direction. And it was from that perspective we started Mozilla Festival sixteen years ago now, which brings together hackers and and all of those other characters I talk about. And and in many ways, that community architecture, things like Mozilla Festival, the fellowships we give for people who are sort of responsible tech, people who are building career around responsible tech. All of those things is is the thing I'm most proud of because we we have, as Mozilla, an amazing community architecture

a whole network of alumni around the world, people who work for Mozilla, people who've been, you know, open source volunteers, people who've been involved in Mozfest. And you really see which is a powerful thing even though you can't always touch it or you can't really get your arms around the the whole thing because it's so massive. People take that ethos, you know, what I think is the ethos in the Mozilla project from, you know, almost thirty years ago, the ethos in the Mozilla manifesto around things like privacy and open source take it into the rest of their worlds, the rest of their careers, the rest of their lives. And in in many ways, while it may not be as tangible as a browser, it is one of the biggest legacies and and the influences of Mozilla.

Alex, it's such a horrible question. I mean, of course, it's the right question. And and I guess, you know, I say I would say in my time caring about these things, working on these things, which is, you know, probably most of my career, the last twenty, thirty years, it's a roller coaster. Yeah. And so, you know, I I started doing open source stuff and web stuff in the mid late nineties. It was a time of tremendous optimism, and it is a thing where this ethos got to be built into product, built into technology, built into some of our institutions in ways that were really liberating. But we've also seen those original web technologies get abused and, you know, get see people get scammed, see us polarize ourselves, all of those things. And and I think we're still on that same roller coaster or maybe it's a dance of a yin yang of these empowering technologies also get used in ways that undermine democracy and and make us feel threatened and anxious as people. So, you know, I I feel that kinda same tension right now as we're in the AI era. And I I think of the AI era as just the next era of digital technology building blocks. The web was one. Personal computers were another. We're just in a different era in terms of what tech we use to build stuff with. And if you asked me couple years ago, I would have been really pessimistic because we had consolidated everything in a few companies, and that still remains the largely the case and the risk. But when OpenAI was first rising, I thought, man, we have just locked this next era technology inside of one company or a couple of companies. And I am more optimistic now and that you you've seen that technology both commoditized, but in in certain regards, democratize much faster than other things. It's like everybody's producing AI models. Their open source is totally caught up or almost completely cut up up to what seemed like these unimpeachable frontier companies. You see stuff coming in in open source AI from, you know, not just The US and China, but lots of parts of the world. So if you look at that macro perspective, I do have some optimism that more people can shape and share in the spoils of with where this technology goes. But then it comes down to what people are really anxious about. What does this stuff actually mean? What are we gonna use it for with each other? Are we gonna take away all of our jobs? Is this stuff gonna dumb us down? Hey. We are right at the very beginning of grappling with those questions. I'm curious how you think about that tension. So oftentimes and this this felt particularly true a couple years ago where

is that tension between if you open something up, there's tremendous opportunity for empowerment, for creativity, for entrepreneurship, for hacking and bending and sharing in the spoils of the resource because I can start something in my little town or my country, wherever that open source came from and and build from there. So, you know, open source has has in it some real potential to to be empowering be because of its its nature, and it has the potential to be abused, to be enclosed, and all of those things. So we, you know, we sit in that balance. And I and I think, generally, in the last twenty years of of open source, enough people with enough skin in the future of where the industry goes, where society goes, have built on open source that they then need to also be the guardians of it. And that's where I think that that first debate around open source AI being riskier than closed AI got it wrong and frankly was was instrumentalized by a certain set of players. So So if we look at the last twenty, thirty years of cybersecurity, the fact that the core infrastructure of the Internet, the Linux stack, all these things is open, means that everybody who relies on it, which is kind of everybody who uses technology, has an interest in it saying safe and secure. So when there's a bug, when there's a critical security vulnerability, bad things can happen, but everybody rushes to solve it. And, you know, that, I think, can be the case if we build this AI era on open source that, of course, problems will come. People will abuse. Bad actors will act badly. But if everybody is on a kind of common open infrastructure, we can all see how it works, and we can all rush in to to try to fix it. And that's the the world that I think is most viable in terms of us balancing kind of empowerment and and safety. And and if you go back a couple years where they're in there, there's still people like this, who are just saying open source is super dangerous in the AI era. Many of those people were the people who really wanna position themselves the arbiters of our safety. And and I the idea that one or two or three companies are the the people we wanna trust with the safety of humanity and this technology, that just feels like a very narrow bet in an era where people don't trust companies or institutions.

there is tremendous potential in this technology that lets us create more easily or lets us automate things more easily. I mean, it it just it done right, like the web, generative AI is a LEGO box that we can all build things with. And so, you know, we want that LEGO box to be diverse and democratized. But also at the same time, I think people are anxious about what it could unleash, and also some people just don't want any of it. So I think at this at the starting point with Firefox, and, you know, some of the things coming out of the Thunderbird team, I think they're they're even more cautious because they're so focused on core privacy audience on on embracing AI. What you see is a little bit more of a slow approach to figure out how do we build AI into our massive consumer products with a lot of focus on choice and control for people. So So if you've been watching Firefox in the last few months, there's a a push to figure out how can AI be integrated into the browser in ways that people will find helpful, desirable, interesting, creative, playful. And I think we'll do some some stuff that's different than the other browsers. You're gonna see it roll out in the next few months. At the same time, there's a real emphasis on you can turn it all off. There's a single button. We originally called it the kill switch, but it's now called AI controls that you say, like, I want all AI features in Firefox turned off. And at the same time also, you can choose from lots of different models based on what you want to do with AI in your browser. It's not like if I go use Chrome, I'm stuck with Gemini. If I go use Edge, I'm stuck with Copilot. You'll be able to to choose from the beginning with the smart window in Firefox a number of models. But over time, just like you can search from dozens of search engines in Firefox and enable whatever you want, that same kind of control over your AI experience. So that's one thread of it. But the other thread is really just making sure that open source has a chance to be the winning paradigm for developers. And increasingly, more and more of us are developers. And so we've spun out separate new companies owned by the foundation, one called Mozilla AI, one called Mozilla Data Collective, and, set up a venture fund, Mozilla Ventures. Mostly, all of that is focused on making it possible that open source is easier to use, is more secure, is more governable in AI. Because the the real challenge with open source AI right now isn't that it isn't capable. It is increasingly capable. And it isn't that it isn't, cheaper. It is cheaper, and so developers, enterprises want it. It's actually that it's harder to use than just going and plugging into an API from Anthropic or an API from OpenAI or or Google. So these companies we've set up and invested in are really focused on making it easier to use, build on, train, customize, open source AI so that that becomes the the mainstream option for people who are building things. So I think those are the two things. It's more choice, more control for people using things using AI through things like Firefox, and then

to make the perfect browser for one particular audience, but to make sure that many people, many audiences, have the browser they want and can be able to shape it and control it. And so I think that's a key piece is that we're not just looking for one consumer base that then we can go and monetize or suck their data. We don't wanna do either of those things. We're looking for saying, I want a low friction experience that just feels easy to use and I trust you. That's an option. I want something where I can turn off all of the AI. You know, that's there too. I'm a developer. I wanna use this tool in a certain way. That's there too. So I think the first thing is to really engage with and listen to different communities that are a part of who uses Firefox, who uses Thunderbird, who builds with us in our open source tools, and adapt with them. Because right now, I don't think any of us really know what the new user behaviors are gonna be. Like, this is, like, 1996 in web browsing land, right, in terms of knowing what was the first generation of the Internet and the second generation four. Things that we're still learning, and our commitment is to kinda listen to people and shape it with them. And then the other thing is to apply those principles in the Mozilla manifesto, which we've done for twenty years in day to day design work. So thinking about, is there a way to make this more private as we implement it, which is not what most others are incented to do. Is there a way to collect less data or no data or let it be thrown away? You know, we were the first mainstream browser to to we're probably the last mainstream browser to include translation, you know, AI based translation, but we're the first to make it based on local models. And we waited till we could get to the point where you could do machine translation effectively in the browser using local models so that information is not transmitted up to the cloud and used by somebody else. And so those are the kinds of design decisions that product managers and engineers in our consumer product groups in Firefox and Thunderbird, like, are tuned to make all the time. I think even in advertising, you know, it it to some, it's maybe even controversial that we've added native advertising into Firefox. But that is built by people who are the best in the industry at thinking about how do you provide ads that are interesting to people, but also do it in a way that isn't sucking everybody's data and that is actually privacy respecting. So you you have engineers and product managers who really are tuned to make decisions with those Mozilla manifesto principles first. There's always trade offs, but I think that's the key. And then you have to remember, we don't know what all those user behaviors are. Hopefully, we, you know, figure out what it is that interest and delights people and is useful to people, and then bring those values to to the table and do stuff that we couldn't even imagine today. I'd love to have you talk a couple more minutes about the data collective project. It's super interesting, really important. Where are you prioritizing? And just explain to the audience kind of what the value proposition is there. So one of the companies that we've spun out and, you know, as I said at the beginning, we're sort of like this set of superheroes where we've we spun out these different companies that are still controlled by the foundation and are, you know, driven by our mission or our manifesto. One of the ones we spun out was recently is called the Mozilla Data Collective. And what the Mozilla Data Collective is focused on is is gonna sound very wonky, but a two sided marketplace for ethical training data. So So what does that mean? Means that, obviously, one of the things we struggle with right now is we're running out of data, which is hard to to imagine. But, you know, all of these models are trained off having sucked everything in public. Whole question there about, you know, is is that right and what do we wanna do about it? But building on AI and doing more useful things requires, at this point, specialized data. I wanna train a small model. I wanna do this for a particular industry or or organization or function. I wanna do this in a particular language. And there's way more data out there that is behind a password than than can be scraped. Thank god. The connection between those two things, you get better results. You do more interesting things with AI if you have specialized data, and that it that data is out there and is protected. We're trying to put those two things together and say, if you are an owner of that data, whether it's cultural information like a language, whether that is some particular insight or kind of academic body of work that you and a community have built up, whether it's something your organization or your company has, you should be able to offer it to people who you trust to use it well in AI applications, but you should be able to do it in your own terms. And so if you think about, you know, what is a two sided marketplace, somebody has something to offer, somebody can come and access it. We wanna let people offer their data for others to use in AI applications or any applications, but mostly it's AI. But to be able to say, this is under this kind of Creative Commons license, or this is only available, you know, for free for people who are in my community but has to be paid for if you're a big tech company, or lots of different things. And no nobody else is out there doing that. You know, there are places where you can go and get boutique data, but it tends to still be pretty extractive. And in many ways, I mean, I wouldn't say we're a union for people who have specialized data, but we really wanna put the rights of data holders first. But then also let, you know, let this stuff be used if people wanna let it be used for interesting applications. And that balance between data holders, I mean, we see this in the whole of society with AI, and data consumers or people using data in AI is really out of whack right now. So this is our one small approach to to kinda try to put it more in whack. And and the more in whack. It's a funny thing to say, but more in balance. And and and the last thing I'll just say is where where we got the idea was we believe that open datasets can be really valuable, and especially open datasets that that bring the diversity of humanity to the fore. And we worked for almost a decade on a project called Common Voice, which is like a Wikipedia for training data for for voice, you know, if I wanna train a voice AI application. And we have the most diverse training data for voice AI in the world, lots of different accents, lots of smaller languages. And and that's valued by itself, and we believe in in continuing to build datasets like that. But a number of communities, including, you know, the Maori community in New Zealand came and said, we don't actually wanna put our stuff out in this open dataset. We wanna make it available on our terms. And so Mozilla Data Collective really comes from that insight is, you know, we want rich, diverse, interesting data out there, but it it shouldn't just be either all commercial, all open source, all anything. We wanna kinda build a fair way for people to negotiate back and forth on this stuff. That's what the Mozilla Data Collective is.

We are stewards of this notion, and it's, you know, it's not me. It's the team behind Convoyce that's then gone out to to build this new enterprise, and they they have been amazing stewards of this notion. And and I wouldn't say we've learned that much yet on how this marketplace that rebalances things works because we've just started. I mean, what we've learned is that nobody is filling that gap. And I guess what we're seeing early on is, you know, governments and communities who hold, you know, kind of public content or cultural content are already being thoughtful about this stuff. So they're often the ones becoming our design projects and some of the the early datasets and features we're rolling into here. But I do think it will you know, what we need to do, like Mozilla has always done, build the product to try out how you live the values and then learn and and iterate. And I we're with the Mozilla Data Collective, we're still really early in that process. I wanna come back to AI and privacy for a minute. We talked about AI and the browser and kind of user controls

there's still there's the deep privacy questions we face as humanity that we shouldn't cede ground on. And then how do they play out in the AI era? And I think some of it just goes back to incentives and and then the technology for privacy. And so having incentives that are not about just total extraction of data, whether that means that, you know, you're able to monetize as we are with our advertising products that that we put in Firefox in ways that aren't about sucking and holding people's data, but you find creative approaches using privacy enhanced technologies, using different business business models, being willing to accept less and not to maximize monetization, or moving into subscription business models where, you know, the the incentive is making a promise that you're just never gonna use the data and and, you know, it's not that the product is your data. The the product is the product and and you pay for it. So I think there's a piece around incentives and business models, which Mozilla has always been good at and people like Proton are as well. Making sure as we move into AI that we have business models that are privacy compatible, whether that is, you know, less revenue maximization or subscriptions or other things. And people often don't think about that when they think about privacy, but it it is really quite critical.

I think so, and we think so. And that doesn't mean and I I think often, especially as, you know, CDT and Mozilla's people in the in the privacy community, that doesn't mean inherently no advertising or zero data. What it means is whatever you're doing, building in privacy as a part of the business model. And that's where I think it is a moment to push ourselves and say, are there ways to actually be more privacy protecting and still monetize whether it is on advertising or subscriptions that are tied to a privacy promise or other things? And and it may be that there are ways to to do this, which are different now because AI itself like, let's say, we had completely local AI on my machine, on my phone, on my laptop. There may be ways because there's so much context in the models themselves to deliver recommendations that are totally ethical. And those recommendations could be for shopping. They could be for advertising. They could be for whatever that I opt into that are things that I want that I can push away where the data never leaves the machine. So, you know, that's I'm not saying we're working on that because we're not. But theoretically, you know, this is a thing, and I think I I certainly hear people talking about how do we use the capabilities and affordances of these new technologies to deliver some of what we want in terms of discovery and and commercial information exchange, which ultimately, like, most people want some of that in their lives. How do we do that more ethically and privacy respecting ways and build a business model on that? And then I I think the other piece then is is on the tech level, whether in that stuff, it is embracing, you know, things like federated learning or differential privacy or other privacy enhanced technologies or just actually making sure that there are products. And this is something you will probably see from one Mozilla organization, I won't say which one yet, that have encryption built into how all of the chatbot offerings work, which is not, I think, the priority or the interest of the big providers just, you know, just as it hasn't been in, in the previous era. And so I I do think, like, encryption, you see it, Moxie Marlins spike from who founded Signal is out there building, Confer, which is an encrypted chatbot. I think Proton has one, Lumo. I think more and more stuff in that class of things because this is such intimate technology. It's just as intimate or more intimate than messaging, which is the place where people have embraced encryption. But we're building up our memories in these things and are using them as sounding boards and using them to plan our businesses. Having stuff that never really gets shared, never actually gets shared, both the vendor of that technology. I know I want more and more of. I use probably 60% of my generative AI chatbot uses in an encrypted chatbot because there's just so much that I don't want any vendor who I would be using to to, to see. And then I I just will say one last thing, which is a little further out. So I think the business model innovation, we're adding inflection point, privacy compatible business models, looking at encryption and privacy oriented technology in the in the context of these technologies. And then the third is really trying to crack some of the bigger things we haven't cracked in the last era. And I think a lot of that is, you know, Tim Berners Lee, people like that have talked about owning your own data. That becomes a much and that's not only about privacy. That is about agency. It's about portability. It's about, a lot of different things. And I would say this is the time to be asking about that. And we are working on some experimental technologies where you could take your AI memory from a chatbot and have it be private, have it be portable. So as I build up my own history and context inside of ChatGPT or inside of some open source encrypted model or some third thing, I can take it with me. And these things get smarter about us as we use them because we build up history, but that is owned in the most part right now by that company. It's like I'm renting, you know, this space and I'm decorating it and I'm making it better. But then if I wanna go somewhere else, I just have to forfeit it. It's like if I, you know, rented an apartment, I couldn't take my furniture when I left. And and so technologies that let me have control over my own data, my own history in AI can let me keep it private in some ways and let me have it be portable, I are really a critical thing right now. And it it is, as I say, it is something that at the CTO level of Mozilla and with some experiments, we're really seeing if we can crack. And that's another great example where actually innovations in the technology make that easier

to be able to run a data extraction process. Yeah. No. I I remember working on, like, weird metadata standards for syndication of content and, you know, on the web, you know, thirty years ago. And just the structured data and the way things were and, like, how we did in interoperability is much harder than potentially now, because you're dealing with with unstructured data. I mean, there's other challenges, but,

what would it be? Can I have two? Go for it. I mean, I I I think there are two and they're quite different. So I do think for the reasons we just discussed, now is the time for governments to mandate interoperability and portability of data. It's gonna be easier. It's gonna be more important. Businesses need it. People need it. And, you know, there's just no reason not to do it now, you know, other than that there's a bunch of companies that it's, you know, against their interests, which I guess is a thing in politics. But I think really mandating data portability and interoperability that now is the time as we move into this new era of the of the digital tech you know, digital technology and and the Internet industry. And, you know, that that just feels so important. And it's also good for competition. Right? I mean, it doesn't matter if you're on what side of the aisle. Like, you competition is something that really matters and and I think to to everybody. And interoperability and portability are a part of that, and that's a much better thing to invest in upfront than some kinda long antitrust process, you know, decades after it matters. Mhmm. And then I guess the other thing is really governments picking open source. I think there's a way to save public dollars. There's a way to have AI that is more secure and trustworthy. And, frankly, there's a way for governments to share the costs amongst each other because you're leveraging a technology stack. And in some ways, as somebody improves one part of it, somebody improves another part, there's a tremendous tremendous amount of cost leverage. And so I would say, in particular, governments who are in parts of The US or or countries, who are trying to kinda get ahead in AI and carve their own niche, you know, really prioritize open source, procure open source, support open source businesses in your jurisdiction. I I think that's a a real key thing for governments to do.

I appreciate the partnership. Thanks, Alex. Thank you for listening to Tech Talks, presented by the Center for Democracy and Technology. I've been your host, Jamal Magdi. Tech Talks is edited by Jacob Kaufman and produced by Drew Courtney. Check out more of CDT's work by visiting us online at cdt.org and on various social media at sendem tech. That's c e n d e m tech. Thanks for talking tech.