Is your face truly your own, or is it a commodity to be sold, a weapon to be used against you? A company called Clearview AI has scraped the internet to gather (without consent) 30 billion images to support a tool that lets users identify people by picture alone. Though it’s primarily used by law enforcement, should we have to worry that the eavesdropper at the next restaurant table, or the creep who’s bothering you in the bar, or the protestor outside the abortion clinic can surreptitiously snap a pic of you, upload it, and use it to identify you, where you live and work, your social media accounts, and more?

Privacy info. This embed will serve content from

Listen on Spotify Podcasts Badge Listen on Apple Podcasts Badge  Subscribe via RSS badge

(You can also find this episode on the Internet Archive and on YouTube.)

New York Times reporter Kashmir Hill has been writing about the intersection of privacy and technology for well over a decade; her book about Clearview AI’s rise and practices was published last fall. She speaks with EFF’s Cindy Cohn and Jason Kelley about how face recognition technology’s rapid evolution may have outpaced ethics and regulations, and where we might go from here. 

In this episode, you’ll learn about: 

  • The difficulty of anticipating how information that you freely share might be used against you as technology advances. 
  • How the all-consuming pursuit of “technical sweetness” — the alluring sensation of neatly and functionally solving a puzzle — can blind tech developers to the implications of that tech’s use. 
  • The racial biases that were built into many face recognition technologies.  
  • How one state's 2008 law has effectively curbed how face recognition technology is used there, perhaps creating a model for other states or Congress to follow. 

Kashmir Hill is a New York Times tech reporter who writes about the unexpected and sometimes ominous ways technology is changing our lives, particularly when it comes to our privacy. Her book, “Your Face Belongs To Us” (2023), details how Clearview AI gave facial recognition to law enforcement, billionaires, and businesses, threatening to end privacy as we know it. She joined The Times in 2019 after having worked at Gizmodo Media Group, Fusion, Forbes Magazine and Above the Law. Her writing has appeared in The New Yorker and The Washington Post. She has degrees from Duke University and New York University, where she studied journalism. 


What do you think of “How to Fix the Internet?” Share your feedback here. 


Madison Square Garden, the big events venue in New York City, installed facial recognition technology in 2018, originally to address security threats. You know, people they were worried about who'd been violent in the stadium before, or Or perhaps the Taylor Swift model of, you know, known stalkers wanting to identify them if they're trying to come into concerts.

But then in the last year, they realized, well, we've got this system set up. This is a great way to keep out our enemies, people that the owner, James Dolan, doesn't like, namely lawyers who work at firms that have sued him and cost him a lot of money.

And I saw this, I actually went to a Rangers game with a banned lawyer and it's, you know, thousands of people streaming into Madison Square Garden. We walk through the door, put our bags down on the security belt, and by the time we go to pick them up, a security guard has approached us and told her she's not welcome in.

And yeah, once you have these systems of surveillance set up, it goes from security threats to just keeping track of people that annoy you. And so that is the challenge of how do we control how these things get used?

That's Kashmir Hill. She's a tech reporter for the New York Times, and she's been writing about the intersection of privacy and technology for well over a decade.

She's even worked with EFF on several projects, including security research into pregnancy tracking apps. But most recently, her work has been around facial recognition and the company Clearview AI.

Last fall, she published a book about Clearview called Your Face Belongs to Us. It's about the rise of facial recognition technology. It’s also about a company that was willing to step way over the line. A line that even the tech giants abided by. And it did so in order to create a facial search engine of millions of innocent people to sell to law enforcement.

I'm Cindy Cohn, the Executive Director of the Electronic Frontier Foundation.

And I'm Jason Kelley, EFF’s Activism Director. This is our podcast series How to Fix the Internet.

The idea behind this show is that we're trying to make our digital lives BETTER. At EFF we spend a lot of time envisioning the ways things can go wrong — and jumping into action to help when things DO go wrong online. But with this show, we're trying to give ourselves a vision of what it means to get it right.

It's easy to talk about facial recognition as leading towards this sci-fi dystopia, but many of us use it in benign - and even helpful - ways every day. Maybe you just used it to unlock your phone before you hit play on this podcast episode.

Most of our listeners probably know that there's a significant difference between the data that's on your phone and the data that Clearview used, which was pulled from the internet, often from places that people didn't expect. Since Kash has written several hundred pages about what Clearview did, we wanted to start with a quick explanation.

Clearview AI scraped billions of photos from the internet -

Billions with a B. Sorry to interrupt you, just to make sure people hear that.

Billions of photos from, the public internet and social media sites like Facebook, Instagram, Venmo, LinkedIn. At the time I first wrote about them in January, 2020, they had 3 billion faces in their database.

They now have 30 billion and they say that they're adding something like 75 million images every day. So a lot of faces, all collected without anyone's consent and, you know, they have paired that with a powerful facial recognition algorithm so that you can take a photo of somebody, you know, upload it to Clearview AI and it will return the other places on the internet where that face appears along with a link to the website where it appears.

So it's a way of finding out who someone is. You know, what their name is, where they live, who their friends are, finding their social media profiles, and even finding photos that they may not know are on the internet, where their name is not linked to the photo but their face is there.


Wow. Obviously that's terrifying, but is there an example you might have of a way that this affects the everyday person. Could you talk about that a little bit?


Yeah, so with a tool like this, um, you know, if you were out at a restaurant, say, and you're having a juicy conversation, whether about your friends or about your work, and it kind of catches the attention of somebody sitting nearby, you assume you're anonymous. With a tool like this, they could take a photo of you, upload it, find out who you are, where you work, and all of a sudden understand the context of the conversation. You know, a person walking out of an abortion clinic, if there's protesters outside, they can take a photo of that person. Now they know who they are and the health services they may have gotten.

I mean, there's all kinds of different ways. You know, you go to a bar and you're talking to somebody. They're a little creepy. You never want to talk to them again. But they take your picture. They find out your name. They look up your social media profiles. They know who you are.
On the other side, you know, I do hear about people who think about this in a positive context, who are using tools like this to research people they meet on dating sites, finding out if they are who they say they are, you know, looking up their photos.

It's complicated, facial recognition technology. There are positive uses, there are negative uses. And right now we're trying to figure out what place this technology should have in our lives and, and how authorities should be able to use it.

Yeah, I think Jason's, like, ‘this is creepy’ is very widely shared, I think, by a lot of people. But you know the name of this is How to Fix the Internet. I would love to hear your thinking about how facial recognition might play a role in our lives if we get it right. Like, what would it look like if we had the kinds of law and policy and technological protections that would turn this tool into something that we would all be pretty psyched about on the main rather than, you know, worried about on the main.

Yeah, I mean, so some activists feel that facial recognition technology should be banned altogether. Evan Greer at Fight for the Future, you know, compares it to nuclear weapons and that there's just too many possible downsides that it's not worth the benefits and it should be banned altogether. I kind of don't think that's likely to happen just because I have talked to so many police officers who really appreciate facial recognition technology, think it's a very powerful tool that when used correctly can be such an important part of their tool set. I just don't see them giving it up.

But when I look at what's happening right now, you have these companies like not just Clearview AI, but PimEyes, Facecheck, Eye-D. There's public face search engines that exist now. While Clearview is limited to police use, these are on the internet. Some are even free, some require a subscription.  And right now in the U. S., we don't have much of a legal infrastructure, certainly at the national level about whether they can do that or not. But there's been a very different approach in Europe where they say, that citizens shouldn't be included in these databases without their consent. And, you know, after I revealed the existence of Clearview AI, privacy regulators in Europe, in Canada, in Australia, investigated Clearview AI and said that what it had done was illegal, that they needed people's consent to put them in the databases.

So that's one way to handle facial recognition technology is you can't just throw everybody's faces into a database and make them searchable, you need to get permission first. And I think that is one effective way of handling it. Privacy regulators actually inspired by Clearview AA actually issued a warning to other AI companies saying, hey, just because there's all these, there's all this information that's public on the internet, it doesn't mean that you're entitled to it. There can still be a personal interest in the data, and you may violate our privacy laws by collecting this information.

We haven't really taken that approach, in the U. S. as much, with the exception of Illinois, which has this really strong law that's relevant to facial recognition technology. When we have gotten privacy laws at the state level, it says you have the right to get out of the databases. So in California, for example, you can go to Clearview AI and say, hey, I want to see my file. And if you don't like what they have on you, you can ask them to delete you. So that's a very different approach, uh, to try to give people some rights over their face. And California also requires that companies say how many of these requests they get per year. And so I looked and in the last two years fewer than a thousand Californians have asked to delete themselves from Clearview's database and you know, California's population is very much bigger than that, I think, you know 34 million people or so and so I'm not sure how effective those laws are at protecting people at large.

Here’s what I hear from that. Our world where we get it right is one where we have a strong legal infrastructure protecting our privacy. But it’s also one where if the police want something, it doesn’t mean that they get it. It’s a world where control of our faces and faceprints rests with us, and any use needs to have our permission. That’s the Illinois law called BIPA - the Biometric Privacy Act, or the foreign regulators you mention.
It also means that a company like Venmo cannot just put our faces onto the public internet, and a company like Clearview cannot just copy them. Neither can happen without our affirmative permission.

I think of technologies like this as needed to have good answers to two questions. Number one, who is the technology serving - who benefits if the technology gets it right? And number two, who is harmed if the technology DOESN’T get it right?

For police use of facial recognition, the answers to both of these questions are bad. Regular people don’t benefit from the police having their faces in what has been called a perpetual line-up. And if the technology doesn’t work, people can pay a very heavy price of being wrongly arrested - as you document in your book, Kash.

But for facial recognition technology allowing me to unlock my phone and manipulate apps like digital credit cards, I benefit by having an easy way to lock and use my phone. And if the technology doesn’t work, I just use my password, so it’s not catastrophic. But how does that compare to your view of a fixed facial recognition world, Kash?

Well, I'm not a policymaker. I am a journalist. So I kind of see my job as, as here's what has happened. Here's how we got here. And here's how different, you know, different people are dealing with it and trying to solve it. One thing that's interesting to me, you brought up Venmo, is that Venmo was one of the very first places that the kind of technical creator of Clearview AI, Hoan Ton-That, one of the first places he talked about getting faces from.

And this was interesting to me as a privacy reporter because I very much remembered this criticism that the privacy community had for Venmo that, you know, when you've signed up for the social payment site, they made everything public by default, all of your transactions, like who you were sending money to.

And there was such a big pushback saying, Hey, you know, people don't realize that you're making this public by default. They don't realize that the whole world can see this. They don't understand, you know, how that could come back to be used against them. And, you know, some of the initial uses were, you know, people who were sending each other Venmo transactions and like putting syringes in it and you know, cannabis leaves and how that got used in criminal trials.

But what was interesting with Clearview is that Venmo actually had this iPhone on their homepage on and they would show real transactions that were happening on the network. And it included people's profile photos and a link to their profile. So Hoan Ton-That sent this scraper to and it would just, he would just hit it every few seconds and pull down the photos and the links to the profile photos and he got, you know, millions of faces this way, and he says he remembered that the privacy people were kind of annoyed about Venmo making everything public, and he said it took them years to change it, though.

We were very upset about this.

Yeah, we had them on our, we had a little list called Fix It Already in 2019. It wasn't a little, it was actually quite long for like kind of major privacy and other problems in tech companies. And the Venmo one was on there, right, in 2019, I think was when we launched it. In 2021, they fixed it, but that was right in between there was right when all that scraping happened.

And Venmo is certainly not alone in terms of forcing everyone to make their profile photos public, you know, Facebook did that as well, but it was interesting when I exposed Clearview AI and said, you know, here are some of the companies that they scraped from Venmo and also Facebook and LinkedIn, Google sent Clearview cease and desist letters and said, Hey, you know, you, you violated our terms of service in collecting this data. We want you to delete it, and people often ask, well, then what happened after that? And as far as I know, Clearview did not change their practices. And these companies never did anything else beyond the cease and desist letters.

You know, they didn't sue Clearview. Um, and so it's clear that the companies alone are not going to be protecting our data, and they've pushed us to, to be more public and now that is kind of coming full circle in a way that I don't think people, when they are putting their photos on the internet were expecting this to happen.

I think we should start from the source, which is, why are they gathering all these faces in the first place, the companies? Why are they urging you to put your face next to your financial transactions? There's no need for your face to be next to a financial transaction, even in social media and other kinds of situations, there's no need for it to be public. People are getting disempowered because there's a lack of privacy protection to begin with, and the companies are taking advantage of that, and then turning around and pretending like they're upset about scraping, which I think is all they did with the Clearview thing.

Like there's problems all the way down here. But I don't think that, from our perspective, the answer isn't to make scraping, which is often over limited, even more limited. The answer is to try to give people back control over these images.

And I get it, I mean, I know why Venmo wants photos. I mean, when I use Venmo and I'm paying someone for the first time, I want to see that this is the face of the person I know before I send it to, you know, @happy, you know, nappy on Venmo. So it's part of the trust, but it does seem like you could have a different architecture. So it doesn't necessarily mean that you're showing your face to the entire, you know, world. Maybe you could just be showing it to the people that you're doing transactions with.

What we were pushing Venmo to do was what you mentioned was make it NOT public by default. And what I think is interesting about that campaign is that at the time, we were worried about one thing, you know, that the ability to sort of comb through these financial transactions and get information from people. We weren't worried about, or at least I don't think we talked much about, the public photos being available. And it's interesting to me that there are so many ways that public defaults, and that privacy settings can impact people that we don't even know about yet, right?

I do think this is one of the biggest challenges for people trying to protect their privacy is, it's so hard to anticipate how information that, you know, kind of freely give at one point might be used against you or weaponized in the future as technology improves.

And so I do think that's really challenging. And I don't think that most people, when they're kind of freely putting Photos on the internet, their face on the internet were anticipating that the internet would be reorganized to be searchable by face.

So that's where I think regulating the use of the information can be very powerful. It's kind of protecting people from the mistakes they've made in the past.

Let’s take a quick moment to say thank you to our sponsor. “How to Fix the Internet” is supported by The Alfred P. Sloan Foundation’s Program in Public Understanding of Science and Technology. Enriching people’s lives through a keener appreciation of our increasingly technological world and portraying the complex humanity of scientists, engineers, and mathematicians. And now back to our conversation with Kashmir Hill.

So a supporter asked a question that I'm curious about too. You dove deep into the people who built these systems, not just the Clearview people, but people before them. And what did you find? Are these like Dr. Evil, evil geniuses who intended to, you know, build a dystopia? Or are there people who were, you know, good folks trying to do good things who either didn't see the consequences of what they're looking at or were surprised at the consequences of what they were building

The book is about Clearview AI, but it's also about all the people that kind of worked to realize facial recognition technology over many decades.
The government was trying to get computers to be able to recognize human faces in Silicon Valley before it was even called Silicon Valley. The CIA was, you know, funding early engineers there to try to do it with those huge computers which, you know, in the early 1960s weren't able to do it very well.

But I kind of like went back and asked people that were working on this for so many years when it was very clunky and it did not work very well, you know, were you thinking about what you are working towards? A kind of a world in which everybody is easily tracked by face, easily recognizable by face. And it was just interesting. I mean, these people working on it in the ‘70s, ‘80s, ‘90s, they just said it was impossible to imagine that because the computers were so bad at it, and we just never really thought that we'd ever reach this place where we are now, where we're basically, like, computers are better at facial recognition than humans.

And so this was really striking to me, that, and I think this happens a lot, where people are working on a technology and they just want to solve that puzzle, you know, complete that technical challenge, and they're not thinking through the implications of what if they're successful. And so this one, a philosopher of science I talked to, Heather Douglas, called this technical sweetness.

I love that term.

This kind of motivation where it's just like, I need to solve this, the kind of Jurassic Park, the Jurassic Park dilemma where it's like,it'd be really cool if we brought the dinosaurs back.

So that was striking to me and all of these people that were working on this, I don't think any of them saw something like Clearview AI coming and when I first heard about Clearview, this startup that had scraped the entire internet and kind of made it searchable by face. I was thinking there must be some, you know, technological mastermind here who was able to do this before the big companies, the Facebooks, the Googles. How did they do it first?

And what I would come to figure out is that. You know, what they did was more of an ethical breakthrough than a technological breakthrough. Companies like Google and Facebook had developed this internally and shockingly, you know, for these companies that have released many kind of unprecedented products, they decided facial recognition technology like this was too much and they held it back and they decided not to release it.

And so Clearview AI was just willing to do what other companies hadn't been willing to do. Which I thought was interesting and part of why I wrote the book is, you know, who are these people and why did they do this? And honestly, they did have, in the early days, some troubling ideas about how to use facial recognition technology.

So one of the first deployments was of, of Clearview AI, before it was called Clearview AI, was at the Deploraball, this kind of inaugural event around Trump becoming president and they were using it because It was going to be this gathering of all these people who had had supported Trump, the kind of MAGA crowd, O=of which some of the Clearview AI founders were part of. And they were worried about being infiltrated by Antifa, which I think is how they pronounce it, and so they wanted to run a background check on ticket buyers and find out whether any of them were from the far left.

And apparently this smartchecker worked for this and they identified two people who kind of were trying to get in who shouldn't have. And I found out about this because they included it in a PowerPoint presentation that they had developed for the Hungarian government. They were trying to pitch Hungary on their product as a means of border control. And so the idea was that you could use this background check product, this facial recognition technology, to keep out people you didn't want coming into the country.

And they said that they had fine tuned it so it would work on people that worked with the Open Society Foundations and George Soros because they knew that Hungary's leader, Viktor Orban, was not a fan of the Soros crowd.

And so for me, I just thought this just seemed kind of alarming that you would use it to identify essentially political dissidents, democracy activists and advocates, that that was kind of where their minds went to for their product when it was very early, basically still at the prototype stage.

I think that it's important to recognize these tools, like many technologies, they're dual use tools, right, and we have to think really hard about how they can be used and create laws and policies around there because I'm not sure that you can use some kind of technological means to make sure only good guys use this tool to do good things and that bad guys don't.

One of the things that you mentioned about sort of government research into facial recognition reminds me that shortly after you put out your first story on Clearview in January of 2020, I think, we put out a website called Who Has Your Face, which we'd been doing research for for, I don't know, four to six months or something before that, that was specifically trying to let people know which government entities had access to your, let's say, DMV photo or your passport photo for facial recognition purposes, and that's one of the great examples, I think, of how sort of like Venmo, you put information somewhere that's, even in this case, required by law, and you don't ever expect that the FBI would be able to run facial recognition on that picture based on like a surveillance photo, for example.

So it makes me think of two things, and one is, you know, as part of the book I was looking back at the history of the US thinking about facial recognition technology and setting up guardrails or for the most part NOT setting up guardrails.

And there was this hearing about it more than a decade ago. I think actually Jen Lynch from the EFF testified at it. And it was like 10 years ago when facial recognition technology was first getting kind of good enough to get deployed. And the FBI was starting to build a facial recognition database and police departments were starting to use these kind of early apps.

It troubles me to think about just knowing the bias problems that facial recognition technology had at that time that they were kind of actively using it. But lawmakers were concerned and they were asking questions about whose photo is going to go in here? And the government representatives who were there, law enforcement, at the time they said, we're only using criminal mugshots.

You know, we're not interested in the goings about of normal Americans. We just want to be able to recognize the faces of people that we know have already had encounters with the law, and we want to be able to keep track of those people. And it was interesting to me because in the years to come, that would change, you know, they started pulling in state driver's license photos in some places, and it, it ended up not just being criminals that were being tracked or people, not always even criminals, just people who've had encounters with law enforcement where they ended up with a mugshot taken.

But that is the the kind of frog boiling of ‘well we'll just start out with some of these photos and then you know we'll actually we'll add in some state driver's license photos and then we'll start using a company called Clearview AI that's scraped the entire internet Um, you know everybody on the planet in this facial recognition database.

So it just speaks to this challenge of controlling it, you know,, this kind of surveillance creep where once you start setting up the system, you just want to pull in more and more data and you want to surveil people in more and more ways.

And you tell some wonderful stories or actually horrific stories in the book about people who were misidentified. And the answer from the technologists is, well, we just need more data then. Right? We need everybody's driver's licenses, not just mugshots. And then that way we eliminate the bias that comes from just using mugshots. Or you tell a story that I often talk about, which is, I believe the Chinese government was having a hard time with its facial recognition, recognizing black faces, and they made some deals in Africa to just wholesale get a bunch of black faces so they could train up on it.

And, you know, to us, talking about bias in a way that doesn't really talk about comprehensive privacy reform and instead talks only about bias ends up in this technological world in which the solution is more people's faces into the system.

And we see this with all sorts of other biometrics where there's bias issues with the training data or the initial data.

Yeah. So this is something, so bias has been a huge problem with facial recognition technology for a long time. And really a big part of the problem was that they were not getting diverse training databases. And, you know, a lot of the people that were working on facial recognition technology were white people, white men, and they would make sure that it worked well on them and the other people they worked with.

And so we had, you know, technologies that just did not work as well on other people. One of those early facial recognition technology companies I talked to who was in business, you know, in 2000, 2001, actually used at the Super Bowl in Tampa in 2000 and in 2001 to secretly scan the faces of football fans looking for pickpockets and ticket scalpers.

That company told me that they had to pull out of a project in South Africa because they found the technology just did not work on people who had darker skin. But the activist community has brought a lot of attention to this issue that there is this problem with bias and the facial recognition vendors have heard it and they have addressed it by creating more diverse training sets.

And so now they are training their algorithms to work on different groups and the technology has improved a lot. It really has been addressed and these algorithms don't have those same kind of issues anymore.

Despite that, you know, the handful of wrongful arrests that I've covered. where, um, people are arrested for the crime of looking like someone else. Uh, they've all involved people who are black. One woman so far, a woman who was eight months pregnant, arrested for carjacking and robbery on a Thursday morning while she was getting her two kids ready for school.

And so, you know, even if you fix the bias problem in the algorithms, you're still going to have the issue of, well, who is this technology deployed on? Who is this used to police? And so yeah, I think it'll still be a problem. And then there's just these bigger questions of the civil liberty questions that still need to be addressed. You know, do we want police using facial recognition technology? And if so, what should the limitations be?

I think, you know, for us in thinking about this, the central issue is who's in charge of the system and who bears the cost if it's wrong. The consequences of a bad match are much more significant than just, oh gosh, the cops for a second thought I was the wrong person. That's not actually how this plays out in people's lives.

I don't think most people who haven't been arrested before realize how traumatic the whole experience can be. You know, I talk about Robert Williams in the book who was arrested after he got home from work, in front of all of his neighbors, in front of his wife and his two young daughters, spent the night in jail, you know, was charged, had to hire a lawyer to defend him.

Same thing, Portia Woodruff, the woman who was pregnant, taken to jail, charged, even though the woman who they were looking for had committed the crime the month before and was not visibly pregnant, I mean it was so clear they had the wrong person. And yet, she had to hire a lawyer, fight the charges, and she wound up in the hospital after being detained all day because she was so stressed out and dehydrated.

And so yeah, when you have people that are relying too heavily on the facial recognition technology and not doing proper investigations, this can have a very harmful effect on, on individual people's lives.

Yeah, I mean, one of my hopes is that when, you know, that those of us who are involved in tech trying to get privacy laws passed and other kinds of things passed can have some knock on effects on trying to make the criminal justice system better. We shouldn't just be coming in and talking about the technological piece, right?

Because it's all a part of a system that itself needs reform. And so I think it's important that we recognize, um, that as well and not just try to extricate the technological piece from the rest of the system and that's why I think EFF's come to the position that governmental use of this is so problematic that it's difficult to imagine a world in which it's fixed.

In terms of talking about laws that have been effective We alluded to it earlier, but Illinois passed this law in 2008, the Biometric Information Privacy Act, rare law that moved faster than the technology.

And it says if you want to use somebody's biometrics, like their face print or their fingerprint to their voice print, You need to get their consent, or as a company, or you'll be fined. And so Madison Square Garden is using facial recognition technology to keep out security threats and lawyers at all of its New York City venues: The Beacon Theater, Radio City Music Hall, Madison Square Garden.

The company also has a theater in Chicago, but they cannot use facial recognition technology to keep out lawyers there because they would need to get their consent to use their biometrics that way. So it is an example of a law that has been quite effective at kind of controlling how the technology is used, maybe keeping it from being used in a way that people find troubling.

I think that's a really important point. I think sometimes people in technology despair that law can really ever do anything, and they think technological solutions are the only ones that really work. And, um, I think it's important to point out that, like, that's not always true. And the other point that you make in your book about this that I really appreciate is the Wiretap Act, right?

Like the reason that a lot of the stuff that we're seeing is visual and not voice, // you can do voice prints too, just like you can do face prints, but we don't see that.

And the reason we don't see that is because we actually have very strong federal and state laws around wiretapping that prevent the collection of this kind of information except in certain circumstances. Now, I would like to see those circumstances expanded, but it still exists. And I think that, you know, kind of recognizing where, you know, that we do have legal structures that have provided us some protection, even as we work to make them better, is kind of an important thing for people who kind of swim in tech to recognize.

Laws work is one of the themes of the book.

Thank you so much, Kash, for joining us. It was really fun to talk about this important topic.

Thanks for having me on. It's great. I really appreciate the work that EFF does and just talking to you all for so many stories. So thank you.

That was a really fun conversation because I loved that book. The story is extremely interesting and I really enjoyed being able to talk to her about the specific issues that sort of we see in this story, which I know we can apply to all kinds of other stories and technical developments and technological advancements that we're thinking about all the time at EFF.

Yeah, I think that it's great to have somebody like Kashmir dive deep into something that we spend a lot of time talking about at EFF and, you know, not just facial recognition, but artificial intelligence and machine learning systems more broadly, and really give us the, the history of it and the story behind it so that we can ground our thinking in more reality. And, you know, it ends up being a rollicking good story.

Yeah, I mean, what surprised me is that I think most of us saw that facial recognition sort of exploded really quickly, but it didn't, actually. A lot of the book, she writes, is about the history of its development and, um, You know, we could have been thinking about how to resolve the potential issues with facial recognition decades ago, but no one sort of expected that this would blow up in the way that it did until it kind of did.

And I really thought it was interesting that her explanation of how it blew up so fast wasn't really a technical development as much as an ethical one.

Yeah, I love that perspective, right?

I mean, it’s a terrible thing, but it is helpful to think about, right?

Yeah, and it reminds me again of the thing that we talk about a lot, which is Larry Lessig's articulation of the kind of four ways that you can control behavior online. There's markets, there's laws, there's norms, and there's architecture. In this system, you know, we had. norms that were driven across.

The thing that Clearview did that she says wasn't a technical breakthrough, it was an ethical breakthrough. I think it points the way towards, you know, where you might need laws.
There's also an architecture piece though. You know, if Venmo hadn't set up its system so that everybody's faces were easily made public and scrapable, you know, that architectural decision could have had a pretty big impact on how vast this company was able to scale and where they could look.

So we've got an architecture piece, we've got a norms piece, we've got a lack of laws piece. It's very clear that a comprehensive privacy law would have been very helpful here.

And then there's the other piece about markets, right? You know, when you're selling into the law enforcement market, which is where Clearview finally found purchase, that's an extremely powerful market. And it ends up distorting the other ones.


Once law enforcement decides they want something, I mean, when I asked Kash, you know, like, what do you think about ideas about banning facial recognition? Uh, she said, well, I think law enforcement really likes it. And so I don't think it'll be banned. And what that tells us is this particular market. can trump all the other pieces, and I think we see that in a lot of the work we do at EFF as well.

You know, we need to carve out a better space such that we can actually say no to law enforcement, rather than, well, if law enforcement wants it, then we're done in terms of things, and I think that's really shown by this story.

Thanks for joining us for this episode of how to fix the internet.
If you have feedback or suggestions, we'd love to hear from you. Visit EFF. org slash podcast and click on listener feedback. While you're there, you can become a member, donate, maybe pick up some merch, and just see what's happening in digital rights this week and every week.

This podcast is licensed Creative Commons Attribution 4.0 International, and includes music licensed Creative Commons Attribution 3.0 Unported by their creators.

In this episode, you heard Cult Orrin by Alex featuring Starfrosh and Jerry Spoon.

And Drops of H2O, The Filtered Water Treatment, by Jay Lang, featuring Airtone.

You can find links to their music in our episode notes, or on our website at

Our theme music is by Nat Keefe of BeatMower with Reed Mathis

How to Fix the Internet is supported by the Alfred P. Sloan Foundation's program in public understanding of science and technology.

We’ll see you next time.

I’m Jason Kelley.

And I’m Cindy Cohn.