Signal is a nonprofit because if we were a for-profit, we would be pushed to make profit in an industry where making profit is diametrically opposed to privacy. And so it is to maintain the integrity of our mission that we need to resist me having a board or a bunch of VC investors who, like, go to Davos. They're at the, you know, whatever 6 a.m. breakfast and they hear about some AI strategy and they're like, 'Come on, we gotta, you know, drop some of this privacy. How about Signal for Business? How about an enterprise Signal? How about Signal for Education that just collects a little bit of data so we can train this LLM and then we can do the whatever large language model. Um, and we can do all of this, you know, because we need to get that revenue in the door because we need a Series A, we need a, you know, we need a Series B, we need this, right?' So the nonprofit structure is not a virtue — virtuous, you know, it's a pain to operate as a nonprofit as an at-scale global tech program platform. It's really hard, but it is a protective layer that allows us to actually honor our mission and do real innovation in an industry where almost no innovation happens, right? To actually think through how do we solve problems of privacy and rewrite, you know, rewrite the stack so we can do what is default surveillance in a privacy-preserving manner, and then as like a keystone species in the tech ecosystem hold that bar even though there's very little incentive for others to do so.
Hi everybody. Welcome to the American Library in Paris. I'm Rachel Donadio. I'm the curator of cultural programs here, and it is a great honor and a pleasure to have Meredith Whittaker, the president of the Signal Foundation, here with us this evening, along with Stéphanie Balme, who will be co-moderating with me this evening. We will be in conversation for about 45 minutes and then we're going to open it up to questions from you in the audience. In the Q&A, you should raise your hand and wait for a mic. I will now introduce our speakers.
Meredith Whittaker is Signal's president and a member of the Signal Foundation board of directors. She has over 17 years of experience in tech spanning industry, academia, and government. Before joining Signal as president, she was a Minderoo research professor at NYU and served as the faculty director of the AI Now Institute, which she co-founded. Her research and scholarly work helped shape global AI policy and shift the public narrative on AI to better recognize the surveillance business practices and concentration of industrial resources that modern AI requires. Prior to NYU, she worked at Google for over a decade, where she led product and engineering teams, founded Google's Open Research Group, and co-founded M-Lab, a globally distributed network measurement platform that now provides the world's largest source of open data on internet performance. She was one of the core organizers pushing back against the company's insufficient response to concerns about AI and its harms, and was a central organizer of the Google walkout. She has advised the White House, the FCC, the City of New York, the European Parliament, and many other governments and civil society organizations on privacy, security, artificial intelligence, internet policy, and measurement. She recently completed a term as senior adviser and AI to the chair at the US Federal Trade Commission. So please join me in giving a huge round of applause for Meredith.
Our co-moderator is Stéphanie Balme, who's a researcher and entrepreneur specializing in the geoeconomics of AI. She's an assistant professor at Sciences Po. She explores AI technologies through the lens of software and critical data studies. She also serves on the board of Probal, an AI company, and she has led initiatives at the intersection of innovation, research, education, and emerging markets. As CEO of Digital Africa, she oversaw a $130 million initiative supporting African startups. She's also co-founded Africa for Tech and led OpenClassrooms' strategic expansion in Africa. Her doctoral research traced the rise of AI through a music recommendation algorithm acquired by Spotify, and she teaches at Sciences Po, NYU, and the Sorbonne, and is currently writing a book on the geoeconomics of AI, describing LLMs as belief-making machines. So a big round of applause also for Stéphanie.
Okay, thank you. Without further ado, how many of you regularly use Signal? How — okay. All right. Well, that is good to know. Maybe more of you will after this conversation. Meredith, I think we should open up with a really basic question, which is why should we be worried about privacy? How do we define it? And really, what sets Signal apart from other apps in this perspective? From this perspective?
Hi, thank you. Um, well, thank you for that generous introduction. I think that bio is just too long. Um, you've done so many things. Yeah, that's what happens when you stay in the same industry for two decades. Um, on this question of, you know, privacy, why should we care? That's a — I get that question, you know, a lot. And I think it's, you know, it speaks to our kind of lack of a good definition of privacy that we continue to have to ask that question in a world where the collateral harms are so clear, right? Of the business model built on surveillance, of the aggregation of data and information, of the claim by the tech industry that they know us better than we know ourselves, that they know us better than our communities, than our, you know, that they — they are the right way to know our world. So, you know, there's something there around how we're defining privacy that it continually escapes our grasp as to why it's important, right? Um, and you know, this is actually something, you know, Stéphanie and I were talking about in the back room, like why is it that, you know, this type of authority is sort of seated and taken for granted, that the idea that the tech industry has a right to define who we are to begin with, right? That their database somewhere is a more objective and accurate description of ourselves and our world than our friends, you know, our families, our lovers, you know, this is — this is where that authority, and it's kind of a Hayek for dummies, right? Which I've — I've talked about, like the only way to know is through this sort of objective data. Um, I'm getting already down into a rabbit hole, but I think kind of moving up, I think to understand the answer to that question we need to focus on, you know, what are the precious things we need privacy to protect and not on data flows and processing and where the servers are located, because that gets us into a really boring conversation that people who have laundry and other duties of care just don't have time for, right? Even as these things are existential to their lives. And you know, we can then get into the enumeration. Right. Privacy is necessary to protect democracy because democracy requires a sort of clear empirical basis for reality and a space for contestation. It's required for intimacy. We can't — you know, intimacy is by definition private. It's a shared experience or communication with some or some people that is not shared beyond that. It is necessary for the, you know, for I would say learning, right? You need to have time to absorb new ideas, let those synthesize that into the bedrock of, you know, existing paradigms and disrupt them. It's a very vulnerable process. It is necessary to be able to whistleblow. It's necessary to be able to challenge the status quo in environments where that could be dangerous, right? In environments where maybe you're that one person at Enron looking at what they're doing with energy markets in California saying, 'Maybe this is wrong. Maybe I need to tell someone about this because this is going to mess people up.' Right? It's necessary in Washington, D.C. to report on what might be happening in ways that might be bad, right? It's necessary in your company. It's necessary to be able to think clearly. Like, raise your hand if you're able to comfortably edit with your boss in the doc. Right. Okay. Now, close your eyes. I'm going to ask you to actually close your eyes because we're actually going to try to do this thing, but it'll be quick. Every single message you've ever written in your lives has just been published on a database somewhere. A link has been sent to everyone you know, everyone you've ever known. Click that link, sir. Your eyes are open. But click that link and you can go there and — okay, there's an LLM that'll summarize this. So, they're not just keyword searching. They can type, 'What has ever been said about me?' They can go through all of that. Everyone, you know, like, you know, you're not a dissident. You're not a whistleblower. But that weird, ranty email you sent when you were really angry, your blood sugar was low. That really mean thing you said about your best friend because you were trying to curry favor with some popular kids. That weird lie you told. The accidental forward of something that you then sort of tried to cover up. All of it, right? Doesn't mean you're a bad person. It means you're human. And we're sort of negotiating these boundaries and these fallibilities as we learn and move through life. It's all there. How many of your relationships would survive? Okay. So, why is privacy important? I open my — I opened my eyes now. Now I've opened my eyes. Okay. You can open your eyes, right? But like I wanted that to hit at the like the emotional and libidinal level. I don't want to talk about data right now. I want to talk about why it matters and then we can talk about data. Okay. So, a large language model. We'll let — yeah, we'll let — it's a — it's a fancy type of AI. It's big. Um, but I will request, sir, and everyone else, that we raise our hands for questions because it can be disruptive to have them just yelled out like that, um, in the flow of a conversation like this. So thank you. Stéphanie, why don't you —
Yeah, I can jump on the — So I was really — Thank you, Meredith, for this beautiful introduction. And I think it's very important indeed to make this concept very, uh, close to us and to understand that because there's a type of like conversation going on around AI security and privacy that seems very either like you need a law degree to, you know, have access to it and we are passively condemned to just like watch it and try to make our opinion out of that. I want to ask you though something very specific because this is something that Signal is very much identified with, which is regarding the backdoors controversies and like why — and also specifically because in France we almost passed a law a couple of months ago, weeks ago, uh, that would have allowed, um, you know, backdoors in encrypted messaging. So backdoors, for those — for not having further people asking, uh, backdoors in technology is a way that you can always like, uh, secure an access with what we call a master key, um, to a system, a hardware, a software, it could be any kind of like technology. And, um, this is, you know, very much, um, controversial because, you know, who can have access to the core system and to what's in it and what's in your phone, what's in your cloud, what's in your, uh, in the software you're using. And so this is something that Signal has been very much identified with because Signal has, uh, pioneered what we called end-to-end encrypted messaging, which is securing the fact that when you are actually texting your best friend to quote Meredith something mean or something nice, nobody else than him or her can actually read it. And this is, um — so this is why backdoors are important in the encrypted messaging, uh, world. So why — why is Signal fighting and how is Signal fighting that? Uh, what are the, you know, what are the incumbents for — for this fight?
Yeah, that's a great question. Um, I might want to go back to some history here because this fight is long. It's, you know, we're fighting a fight that we probably will never be allowed to stop fighting because ultimately this is a struggle for power, not a misunderstanding of the laws of math or how technology works. And it's particularly a struggle for the type of power that constitutes itself via information asymmetry. You know, the kind of power that where you know more and more and more about the people and things you want to dominate as a form of domination, right? And domination doesn't always have to have the ugly edge, right? It's, you know, your employers, the government, etc. To govern, you have more and more data and that's seen as an imperative, you know, of the Weberian state and going forward, what have you. Um, in the 1990s, and I don't know how many of you are sort of internet people who experienced this or have read about it — I was not, I was a child, but I've read about it. Um, you know, you had this moment of the commercialization of networked computation. So the internet came into being. It was not sort of deus ex machina. There were, you know, networks already laid down in the US. These were networked computation that was largely funded by the military and then sort of, you know, kind of positioned around research and academic institutions. So this was the providence of governance, governments and universities. In, you know, toward the end of the 1980s, the beginning of the 1990s, it became clear that databases were getting large enough and more, you know, powerful enough, that networks were getting cheap enough and powerful enough, that a number of technological factors were converging to make the dream of a commercial internet, of commercial networked computation, possible. And so this pushed the issue of privacy and encryption. And encryption, for those of you who aren't, you know, in the same world I am, it is the name for the technology that allows us to authenticate and provide privacy via network communication. You know, encryption is the thing we have to do that, right? And it comes in a number of different forms. Um, and so, you know, up until the end of the 1990s, actually, encryption had been, you know, basically the domain of governments. It was highly regulated, you know, under arms control treaties, right? So, you know, if you were to publish your encryption software, as someone like Dan Bernstein tried to do, and it was downloaded by someone in Europe, you were violating arms control treaties, which kind of doesn't work when you're talking about pollinating the world with a, you know, US model of networked commercial computation because ultimately you have to be able to access that endpoint across the world if you're going to do that. And if that's secured with encryption and you're violating arms control treaties, etc., etc. Nor does it work with a model of commercial networked computation, of this sort of commercial internet that would have anything to do with commerce or anything to do with confidential information because of course you have to keep, you know, commercial transactions private. You need to keep, you know, if you're a business running on this, you need to keep your private stuff private, right? So, there were two forces at odds in the 1990s. It was this sort of old-school force of governments who wanted to keep the encryption to themselves so they could secure what they had but have a backdoor into everyone else. And this new force of the sort of Clinton-era neoliberal zeitgeist for this commercial internet that was like, 'Yo, trillions of dollars are riding on this working for us. So you need to figure out a plan here.' And so this was the cauldron within which something called the Crypto Wars happened. And the Crypto Wars were these sort of policy debates and legal fights and all sorts of things that in 1999 basically resulted — or did result — in the liberalization of encryption. So, you know, the arms — you know, basically there it was no longer controlled by these strict arms control regulations. You know, all but, you know, almost not at all controlled by that. Um, and anyone could develop and deploy very strong encryption anywhere they wanted. Full stop. Right. This, you know, and so we kind of ended that, but there was then this persistent desire to create backdoors. So to ultimately undermine the encryption that was out there, not by regulatory decree but through, you know, different laws or — the desire to continually undermine encryption persisted past the 1990s. Um, as a business model that was built on mass surveillance kind of metastasized. Um, and the truth about, you know, undermining encryption is either it is undermined for everyone. So a backdoor, you know, it undermines your best friend and it undermines your worst enemy, or it works for everyone, which means it protects your worst enemy. And that's just the mathematical reality of encryption. Um, what — you know, I'll stop there because there's a lot more.
If I may just wrap up and also like, um, it's interesting because, you know, we have this conception of like legitimate surveillance, you know, like who is in charge, who can be in charge of like actually watching us. You know, usually it depends on like history, it depends on different countries. And, you know, in France we have a certain way to think of like state surveillance which is very different from in the US. Um, and what I find interesting is that with the backdoor controversy, um, you globalize that discussion, you globalize that question and you bring, you know, a sort of like universal claim to the question of privacy that we were relating, um, earlier. And so that's interesting because what's happening right now in the US, from what we see, um, is that usually people who would be against state surveillance would be libertarians. You know, those people who believe that everything has to be based on, you know, personal freedom and individual freedom, and, you know, state surveillance would be fought back by these people. And those same people who were fighting and who are hardcore libertarians like Elon Musk or Peter Thiel are now in charge of what's becoming a very authoritarian state. So I think that's interesting and I would love to hear your voice on this, which is how do we, you know, like how does this first — did you see this shift happening? You know, like so we were mentioning the 80s and the 90s and the sort of like pragmatic realpolitik, um, that happened. And would you say that there's a shift now, another shift happening that, um, blurred the fault lines for understanding tech and policy? And to what extent how does this impact, um, you know, global organizations like Signal?
Yeah. Um, you know, I don't — ultimately Signal is not a political actor. We're a, you know, a platform that does one thing and we do it really, really well and we stay in our lane, which is uncommon in tech, I know. Um, you know, we provide the gold standard for secure communications and we are open source, we are verifiable, we are, you know, and that's the thing we do. You know, people want us to do something else, no, we do that. Um, and so I think, you know, when I talk about this sort of, you know, the fluxes and changes in tech, I've been in tech almost 20 years, it's changed a lot, you know, and ideological consistency has never been one of the strong suits of its, you know, of its godheads, right? Like I don't — I don't even know that I would characterize all of this in the same way. I don't, you know, I don't know that you ask one libertarian what their views is, it would map to another libertarian's. You know, we're, you know, it's a hodgepodge. Um, you know, so what I will say is, you know, there's a whole debate at the kind of early internet. You can look at like John Perry Barlow's — what was it called? — this Declaration of Independence in Cyberspace or these other things that was sort of pressing and celebrating this idea of the end of sovereignty, the end of the, you know, the nation state and the sort of growth of this borderless, amorphous, freedom-loving thing called cyberspace, right? And so that zeitgeist has been around for a while, you know. Now the idea that, you know, the idea that that would sort of lead to kind of universal liberation has obviously like tarnished pretty badly since then. The idea that somehow cyberspace was ephemeral and not, you know, tethered to the material reality of who paid the electric bills for the data centers, you know, who was actually in charge of updating the software running on those servers, all of that was, you know, has come home to haunt that vision pretty significantly. Um, but when it does come to things like encryption, like these, you know, promises of privacy, these all really are all or nothing in some sense, right? And, you know, when we're talking about network systems, so I can't have encryption and talk to you if you don't have encryption. And if you live in France and I live in California, that remains true. If I move to Barbados, that remains true. The system needs to be consistent, it needs to be interoperable, it needs to be running the same standards and the, you know, software or it will not work to purpose. And that, you know, in a sense what we're talking about when we ask these questions is a core struggle of governance, a core struggle of the sort of, you know, I think persistence of the nation state as a sovereign form in the face of a, you know, a type of power and a type of sort of both diffuse and concentrated power that does operate through these networked infrastructures and that does claim authority based on its sort of operational capacity through these infrastructures. So did that get at your question?
Yes, absolutely. And I want to ask you, because we mentioned that before, to tell us and tell the audience a little bit about Salt Typhoon, um, because I think that's a very good example of like what are all the paradoxes that exist inside, you know, backdoors and the possibility to have backdoors.
Yeah. How many people are familiar with Salt Typhoon? Usually it's almost no one. And so congratulations on the — on the few. Um, and it's almost no one because it wasn't widely reported even though it was fairly catastrophic. Um, but it's a very good example of this sort of, you know, what I would have called the magical thinking at the heart of this desire to, you know, have privacy and encryption for me that many governments have. You know, our intelligence services should have it, our politicians should have it, but not for the — the bad guys should not have it, right? And we want some magical key. I think James Comey in 2015 called it a golden key. We don't want a backdoor, we want a golden key. Um, so, you know, it almost does take on the shape of actual magical thinking, like wizard hats and stuff. Um, and Salt Typhoon was revealed at the end of 2024 as a catastrophic exploitation of interception points or backdoors in the US telecommunications infrastructure. And these are, you know, interception points. So, you know, government access to communications that were mandated by the US government, you know, because they wanted to get to the bad guys and bad things would happen if they didn't. Now, it was — it looks like Chinese, you know, it was hackers affiliated with the Chinese state that exploited these. And it's the kind of thing that you would assume the people who got really upset about Snowden would also be equally upset about, right? Because it was that kind of breach. Millions and millions of, you know, people in the US had their data exploited. This was their SMS data, so, you know, text messages. So imagine that database, right? This was their call data all in the clear. So anything they said just swooped up, saved location data, you know, other intimate telemetry data, all collected, you know, in and around DC and nationwide. And it was revealed that, you know, the Trump administration, the Harris administ — Trump campaign, Harris campaign, um, JD Vance, others had, you know, this type of granular information. So you can, you know, like it's your text messages. It's when you started visiting an oncologist's office, when you started talking to a lawyer, when you started talking to a divorce lawyer, when you started going to, say, a reproductive health clinic, when you started talking to a union organizer, when you started talking to another manager in your department about potentially changing jobs, when you, as potentially an agent of, you know, that was working for the US government, um, secretly even though you were allied with the Chinese state, started walking into the FBI office for meetings. Right? I'm making this up, but I'm making this up because I want it to illustrate that SMS data sounds obscure, as we started with, right? But the consequences of this kind of, you know, attack on these backdoors was profound. And I think one of the reasons we haven't heard that much about it, you know, beyond the fact that like it's not very flattering for the government or the telcos, is that it's very hard to know once you have hackers sort of infiltrated in your systems whether they're actually all gone. So we should assume that a system exploited like this is, you know, continues to be exploited. It's a difficult thing. And then there you're defunding CISA and a lot of the, you know, the people and institutions that would be looking at this. So that's, you know, that is a lesson of like you can't have it for me without also offering it for thee, and the consequences are significant when this — when, you know, technology is the nervous system of our lives and institutions. So it's, you know, there we have it, Salt Typhoon.
Yeah. One of the things that obviously happened recently that made a lot of people who hadn't heard of Signal hear of Signal is, of course, when the editor-in-chief of The Atlantic, Jeffrey Goldberg, discovered that he was added to a Signal thread with much of the Trump cabinet planning a hit on a Houthi target in Yemen. And I just wonder what your takeaways were from that incident and how, if at all, it affected how Signal operates.
Um, I'm not familiar. Um, I mean, I had to read the article twice. I didn't, you know, it was a — it was a soap opera. It was a soap opera. It was a wild soap opera. What I didn't expect was that Signal would become a main character, right? We are a platform, you know, everyone with confidential information they have to share or duty of care uses, right? You know, it is — we hear it from everyone. We talk to any government, any intelligence agency, any boardroom, any CEO. They got their Signal contacts. They're having conversations there because at some point someone was like, 'Yo, put that somewhere private.' And they did, right? It is, you know, is the only mass private communications app that works at that scale that crosses institutional boundaries, that crosses jurisdictional boundaries. That means you can't just, you know, if you're in the US military, it's not just talking to your generals on a red phone. You can also talk to your counterpart in other militaries. You can also talk to your babysitter saying, 'Hey, meet me here,' after you told your general, 'Hey, meet me here.' Right? It eliminates that kind of gap. Now, you know, the story was pretty absurd. But I, you know, the thing about Signal is like we know what was in print. We don't know anything else because we have rearchitected the stack up and down to mean we literally don't have that data. So I, you know, like my comment on that was like, 'Wow, wild boys, what are you doing?' But other than that, I'm a civilian. I'm looking in like everyone else because we have intentionally blinkered ourselves from having any additional information that those using Signal don't have themselves and don't tell us, right, or tell Jeff.
It is a new moment where it — I think in the past we wouldn't have said, 'Oh my god, they used the phone to communicate,' but now it's — Signal does become part of the story. Signal is a nonprofit in a world of gigantic capitalistic megatech companies. Tell us a little bit about why it's important for Signal to be a nonprofit and how exactly that functions at scale.
Yeah, I love this question. Um, we are a nonprofit. Um, and we're not a nonprofit because we think the nonprofit form is particularly virtuous, right? We didn't like choose this because I love fundraising dinners, right? Um, we are in an industry, you know, thanks to this '90s history in part and the decision by the Clinton administration in 1996 when they wrote the rules of the road for the commercial internet to do two things that are in combination very toxic. One, zero restrictions on surveillance. So no privacy law, and there in the US there's still no federal privacy law. And two, to endorse advertising as the economic engine of the internet. So why do those two things matter? Well, the one lets you collect the data and the second lets you — incentivizes you to collect more and more and more data because know your customer, and that's the data that creates the models of types of people that are then sold to advertisers or whoever who bid to reach those types of people and then place targeted ads in our feeds that, you know, they want us to see because that's how they make money when we click. So they incentivize engagement, etc., etc., and the collection of —
more and more and more data and data broker industry and like our credit card data and then, you know, the authority to know our world has been seeded, right? So that is the kind of nucleus of what we're dealing with here. And why does that matter to Signal and why does that matter to the question of us being a nonprofit? Well, the economic engine there is the collection and monetization of data. It's surveillance. So either you collect and monetize it or you build goods and services to do that. And as like AI is built on top of this, right? They only realized these old deep learning algorithms worked when they combined them with all this surveillance data. They were like, 'Oh, the 1980s has something to teach the 2010s because now we can do this for optimization in a way we never could before the surveillance was in place.' So anyway, Signal is a nonprofit because if we were a for-profit, we would be pushed to make profit in an industry where making profit is diametrically opposed to privacy. And so it is to maintain the integrity of our mission that we need to resist me having a board or a bunch of VC investors who like go to Davos. They're at the, you know, whatever 6 a.m. breakfast and they hear about some AI strategy and they're like, 'Come on, we got to, you know, drop some of this privacy. How about Signal for business? How about an enterprise Signal? How about Signal for education that just collects a little bit of data so we can train this LLM and then we can do the whatever large language model.' Um, and we can do all of this, you know, because we need to get that revenue in the door because we need a Series A, we need a, you know, we need a Series B, we need this, right? So the nonprofit structure is not a virtue virtuous, you know, it's a pain to operate as a nonprofit as an at-scale global tech program platform. It's really hard, but it is a protective layer that allows us to actually honor our mission and do real innovation in an industry where almost no innovation happens, right? To actually think through how do we solve problems of privacy and rewrite, you know, rewrite the stack so we can do what is default surveillance in a privacy-preserving manner and then as like a keystone species in the tech ecosystem hold that bar even though there's very little incentive for others to do so.
Well, you're a nonprofit and you're also open source.
Yeah. And we have to be open source if we have any integrity or dignity because, you know, we do not expect you to take my word for it, to take our word for it when the stakes are this high. So the the the code is open so many eyes can scrutinize it, find problems, find issues. If there's someone trying to insert a backdoor, they can call it out, call us out, and make sure it's stopped.
Yes. The founder of Mozilla, Mitchell Baker, calls open source an ethos. She says it's an architecture of participation. Would you agree with that?
I mean, it can be, but there's a bunch of dead repos on GitHub that no one touches, right? Like it's also labor. It's also, you know, it's a lot of things. There's open source is a licensing paradigm, right? It's a legalistic paradigm that, you know, you can spend thousands of messages on a mailing list like turning over. It is it is a part, you know, it's a community participation ethos. So it can be something where many people contribute to something, but a lot of times that isn't the case because that misses a labor politics there. Like people spend their time and energy, there has to be an incentive there. It is also a transparency mechanism. So in Signal's case, we are not a community-driven project in so far as we rarely accept commits from outside because we don't have the resources to thoroughly review thousands of people tossing code over the wall. Right? Sometimes we do, sometimes it's good, sometimes it's great, but we do deeply welcome people commenting on our code, sending in bugs, sending in issues, saying, 'Hey, this looks like a problem,' right? Because we want that scrutiny. So we are open source in the, you know, as a licensing, you can copy our code, you can use it under GPL, and we're open source as a sort of transparency mechanism because the promises we make are we believe so important that we want people to be able to verify them, not simply trust us.
Yeah. That seems like a very different approach than the one of Google where you spent many years and you were involved in the walkout in 2018 to push back against business practices and ethical practices that you disagreed with. You've called it metastatic capitalism and you've also called yourself a dissenting court jester whose critiques enhanced Google's status as a place that accepted dissent but where you felt you ultimately had no power to affect change. And it seems that Google is such an important company today and I wonder if you could tell us a little bit about your experience there, your critique of it and also just why its corporate culture is important for us to know. I mean the corporate culture of Google, we should know about that because it is one of the companies that really is so dominant in the world today. So could you tell us a little bit about that world?
Yeah. Well, look, I left Google July 2019, so about six years ago. So, I sort of can't speak to that culture anymore because that's a long time in tech. You know, it was I joined in 2006, I left in '19. So, you know, I saw some. But I don't think it's actually I just want to be there's like an image of someone who just like banks on credit they earned like decades ago that I just don't want to be. Um, and I don't know that I, you know, what I can say is tech has only grown in terms of its power. They've, you know, hired a lot of McKinsey people. There's now a self-awareness of just how significant the power that has accrued through platform reach, economies of scale, network effects, epistemic control over our information ecosystem and the imbrication of the state and the tech industry that I haven't seen before. Right? I think there is a, you know, that was always true and that was one of the things that has, you know, animated my work and, you know, the the organizing I did at Google, the scholarship I've done is sort of a concern over the way this power was being accrued. And if you, you know, look to history, that concern is very warranted. If you look to the present now, that concern is very warranted. Um, what I hadn't seen before was a sort of self-awareness of exactly what that power meant and how it could be wielded. And and I see that now and I see that being naturalized in a really frightening form with the AI narrative, right? This idea that, um, you know, we are creating godheads that are more intelligent than we are, that we need to trust with our decision making, with our allocations, with our, you know, governmental infrastructure, with, you know, everything from teaching our children to deciding what we eat to monitoring our steps to, you know, whatever this hyper-tailored individuality that is sort of emerging through this, you know, this this paradigm. You know, that's a story we tell, right? It's not real about these technologies, about the weird data they're trained on and the strange inference they do and the amount of compute they use to do all of this. And it's a story that is ultimately crediting a handful of companies that have the resources to build these technologies with godhead intelligence and and ceding the power to make these determinations to the handful of people who ultimately benefit from and and control the technologies at these companies. So it's the the kind of narrative that I see wrapped around this surveillance business model which is now the basis of this kind of AI story that is the most concerning shift between 2019 where this was well on the way, right? We're talking, you know, this AI boom that we're in started like 2012 is the inflection point we can mark. It was sort of a little before then, but 2015 is when all the money hit, the DeepMind acquisition, like DeepFace was 2014. And so we saw this sort of supercharged through that period, you know, GPT is a kind of novel manifestation. But I think the the narrative armor they have to justify doing whatever they want with plausible deniability that they, you know, we don't know where the orders came from, we just know they're right and you need to follow them is what I'm most concerned of here.
