The future of privacy in the AI age, with Signal President Meredith Whittaker

Meredith Whittaker’s path to tech leadership was nontraditional. She was not focused on computers as a kid, and in college, she studied English and not computer science. She never imagined architecting the future through technologies inspired by her favorite science fiction novels, but in 2006, she took a temp job at Google that eventually led to her founding an open research group. 

Near the end of her 13-year stretch working for Google, Whittaker led protests of the company’s military contracts, arguing that Google was putting short-term revenue gains ahead of privacy concerns. After a stint advising the Biden administration on AI policy, she was selected to lead an influential tech organization of her own: the Signal Foundation, the nonprofit behind the privacy-focused Signal messaging app, which has an estimated 100 million users worldwide, including, famously, senior White House officials. 

Signal’s encrypted messaging app is free to use and does not include ads, standing in contrast to many of its competitors in the space. As a nonprofit that relies on grants, donations, and seed money, the Signal Foundation does not do much traditional “business” at all, freeing Whittaker to pursue her true passion: advocating for the privacy rights of all technology users.

In this conversation, Whittaker shares why she thinks privacy is a fundamental human right, how she believes tech’s business models laid the foundation for today’s AI tools, and the challenges she sees governments facing in regulating these technologies.

This interview has been edited for length and clarity.

Namir Khaliq: How do you think about your position at this moment? What role does a privacy expert and head of a nonprofit messaging app have to play in the age of AI?

Meredith Whittaker: I see the experience and knowledge I have gained over 20 years in this strange industry as a way to be an honest broker around these technologies. I can explain how they work. I am happy to discuss their limitations and their possibilities. And I firmly believe that the questions we ask should be based on a discussion of the world we want to build and how, if at all, certain technologies fit into that world.

Part of what I am animated by is the idea of getting back to the fundamentals of, yes, we have large-scale, deep-learning-based AI that has been built on a very particular business model of economies of scale, network effects, mass surveillance, and a consolidation of infrastructural power in the hands of a handful of companies. OK, that’s where AI is coming from. Now, how could it fit into a world in which everyone has the ability to have a joyful life? That is the fundamental question that I am interested in answering. I think privacy as a fundamental human right is a bedrock condition for creating a world in which we can live and grow and transform. 

Khaliq: I’m a member of Gen Z, and it seems to me that, relative to previous generations, my peers don’t value digital privacy much. When it comes to the technologies they use, I think privacy would rank close to the bottom of their concerns, if it made the list at all. 

That may be because we grew up in a world where, for better or worse, a lot of what we’ve done has been recorded online in some form, and yet most of us haven’t seen any grossly negative impacts on our lives from that. Have you noticed this trend as well, and if so, does it concern you?

Whittaker: I think we have to be really careful here because there’s a narrative that has been built up that equates dependency or use of technological products with acquiescence and even acceptance. 

You can say people don’t care about privacy because they’re clicking the terms of service, or they click through cookie banners, or they use Instagram to post things and to communicate with friends. I don’t think we can actually make that case clearly. I don’t know that using Instagram means you’re OK with Meta having your data. I think using Instagram means that you are a human being who wants to participate in life, and Instagram is where your friends post updates. There is a social context that allows you to belong, to be included, to be invited to the parties, to know what people are talking about when you show up for school or work. 

So when I think about whether Gen Z or anyone else cares about privacy, I think we have to be very careful with a statement like “Gen Z uses it, so they’re OK with it.” There are many costs to participating in a full life that we may not be OK with, but they are the parameters of the world we were born into, and we don’t have the individual power to just pull a lever and change them. 

I would say, alongside that, there is also a growing dislike of AI, particularly in the U.S. context, from Gen Z: “AI is corny. AI is creepy. AI is not cool. Everyone hot, everyone creative, they fucking hate AI.” That also has to be put in a basket alongside this idea that Gen Z is placidly OK with an invasion of privacy. [Editor’s note: A 2026 Gallup survey of Gen Zers in the U.S. found that 51% use AI at least weekly, though the generation’s negative sentiment toward the technology increased from 2025 to 2026: less hopeful and excited, more angry and anxious.]

Khaliq: In which ways are the privacy challenges of AI the same as those for other tech categories, like messaging and social media? And in which ways are they unique?

Whittaker: We have to go back into the recent history of AI and recognize that they actually aren’t separable categories. AI isn’t a new thing that came sequentially after social media or after messaging. It’s built on top of a very particular business model and on top of those previous technologies. 

We can see the reemergence of a deep-learning-based approach to AI in the early 2010s. And one of the reasons that it reemerged, after being considered a dead end scientifically for the previous decade, was the platform business model, which involved collecting huge amounts of data. You have Facebook, you have YouTube, you have Gmail, etc., spinning up server infrastructure, data centers, and compute in order to collect, manage, and index huge amounts of data on users, such that you could create profiles that you then sell to advertisers. That is the surveillance advertising business model. 

The infrastructures needed for AI existed already as core parts of this social media business model, which included messaging. It was those resources that actually reanimated interest in deep learning. This is not a new technology. What is new is the powerful compute required to train these models and the huge amounts of data collected via the surveillance business model. In the early 2010s, these companies started recognizing that “Hey, we can use these old deep-learning algorithms with our data and compute.” 

You have to recognize these as co-constituted. We wouldn’t be talking about AI right now if novel ingredients hadn’t emerged out of the surveillance business model, and those ingredients are massive amounts of data — which I will refer to as “surveillance” because that’s a technical term — and huge amounts of computational infrastructure. They were augmented and extended to support deep learning in the hands of the companies that had already developed economies of scale to ensure their monopolistic and entrenched status as dominant players. 

Khaliq: What would be a realistic worst-case scenario for a loss of privacy in the age of AI? What would the nightmare situation look like from your perspective?

Whittaker: AI is a technology that is dependent on the loss of privacy. It requires huge amounts of data to train and inform. It is a product of surveillance — in the form of social media data, in the form of the data that we may have innocently deposited on the web — that is now being used in ways that we never imagined at the time. 

I think the worst-case scenario is that these infrastructures, which include massive amounts of surveillance, are controlled by a handful of companies based in the U.S. and expanded to be the nervous system of governments, economies, and institutions that have the capacity to collect and create more data about us, and that the information environments now supercharged by AI exert extraordinary influence over the global information commons, including what we know, what we don’t know, what we can say, and what we can’t say. 

If these technologies fall into the hands of an authoritarian state or a nexus of entities that are willing to use them in a way that advances agendas of social control, oppression, and domination, I think that is the nightmare scenario. All the pieces are already in place. [The question] is who wields them and to what end? Access to AI infrastructures gives them a much more finely grained ability to shape what we do and don’t know, particularly as we begin to rely on these chatbots as arbiters of information.

Khaliq: Given that the genie is already out of the bottle and seemingly can’t be put back in, what does the best path forward look like? 

Whittaker: I think there are real issues with the limits of this deep-learning-based approach to AI. It’s good at coding. It’s good at math to various degrees. But it is much worse at a kind of critical thinking that it’s also claiming to replace. I think there is a need to sort out the wheat from the chaff in terms of where AI actually belongs. 

Then I think there are core concerns around the way that AI is being implemented — not only the security and privacy concerns of AI models themselves, which are inherently vulnerable, but also the data access requirements and [how] these AI models are being integrated into systems in ways that are incredibly insecure. 

I’m seeing a pretty resounding backlash against AI: from [opposition to] data centers to Gen Z really hating AI slop to AI being used as an epithet to mean “fake” or “corny.” I think these are the ingredients to make meaningful changes to how we relate to these technologies overall. There’s a very powerful combined self-interest of almost everyone who isn’t invested in this handful of massive companies based in the U.S. Most of the world has a self-interest in pushing back against this particular form of AI, given the dangerous dependencies that it creates and the very real concerns around cybersecurity, privacy, sovereignty, and other existential issues. 

Khaliq: How have state relationships with AI technology differed around the world? Which nation or nations do you think have gotten that relationship most right?

Whittaker: What we are looking at when we talk about these AI technologies — if we want to talk about the ability to develop, maintain, and distribute AI — is that only a handful of companies have all of those, and they are all based in the U.S. or China.

AI is fundamentally a technology that is emerging from the U.S. and dependent on U.S. tech companies. What we are seeing now is a real battle around how you could even regulate AI, with people who are threatening meaningful regulations on AI companies being penalized or even persecuted. I would look at the way [former European Commissioner for Internal Market] Thierry Breton was attacked by the U.S. government. He’s been a champion of hard regulation from the European side on U.S. tech companies, and that was seen as censorious interference in the U.S. 

So, it’s not an issue of which countries have gotten this right. The struggle right now is: How could you put guardrails on them, given that you have a U.S. state that appears willing to use its power to defend the companies and use their capabilities for geopolitical advantage? 

Khaliq: In what ways would you argue the U.S. is set up to succeed or fail in matters of privacy and tech? And why do you think that?

Whittaker: I think Edward Snowden showed that we already were set up to fail. These large companies on which AI is dependent, which now own the AI infrastructures globally, are predicated on the surveillance business model. They collect huge amounts of data, and that’s the economic engine. In the late 2000s, we saw the U.S. state availing itself of those surveillance infrastructures to gain geopolitical advantage. 

What we have is a very wicked situation in which there is a shared incentive on the part of these big tech companies, for whom data and surveillance are the economic lifeblood, and a U.S. state that recognizes access to information over its global competitors, over allies and adversaries, over its citizens, and over global populations as a geopolitical advantage, as something that it desires more of, as something that will help it gain leverage, whether it’s with climate negotiations or trade talks or the ability to repress dissent. 

You do have people like Senator [Ron] Wyden who continue pushing [for] privacy and anti-surveillance. But what you don’t have is any real teeth behind any of those ideas or even the possibility of passing a bill that would meaningfully hamper the surveillance business model on which AI is based. I would say we’re even less likely to see a meaningful privacy law coming out of the government now that AI is seen as a national geopolitical imperative to beat China in the AI arms race.

Khaliq: In your ideal world, what should the breakdown of responsibility for harm look like between the major AI labs, the government, and individual users? At which points should the burden of accountability fall onto each entity, and why?

Whittaker: In general, I think the actor with the most power, who’s making the most money, should bear the most responsibility and accountability. Those who are left most vulnerable to harm should have the most leverage, the most power, and the most access to ensure that their perspective is heard, such that they aren’t discounted or seen as collateral damage as the technology proceeds.