 I am TJ Dunovan, Vermont's Attorney General. Thank you, Tish James. Thank you, Carl Racine, on off, on off. That concludes my remarks. Hey, I wanna welcome everybody to the Queen City, Burlington, Vermont. It's taken me 47 years, but I now know how it feels to be cool and hip, because let me tell you something, last night, everybody kept on coming up to me. First of all, let me just point out that across the lake is Tish James Jurisdiction. That's New York. That's not Vermont. General James, thanks for being here. Thank you, General. But everybody kept on coming up to you last night, saying, where do we go to dinner? We can't get a reservation. I felt like I was in the coolest, hottest, hippest city in the world, given recommendations for folks to go. My list, here's the only problem. I haven't been out in 18 months, so I don't know where to go anymore. So I hope folks had a wonderful night. I do wanna welcome everybody here to Burlington for the NAG Conference on the Surveillance Economy, and I wanna thank Chris Toth and the team at NAG for putting this conference together. In particular, I wanna give a shout out to Emily Parsons from NAG for her hard work in organizing this. And I wanna thank, in advance, the hardest-working people in the room, and that is the Hilton staff who are gonna help us throughout these next two days. So thank you. I wanna thank my team, in particular, Charity Clark, my chief of staff, and others for helping to organize this. This was their idea, and if it was a good idea, it's their idea. If it was a bad idea, it's mine. So thank you, Charity, and the folks from the Vermont Attorney General's Office. And I wanna thank my friends and colleagues who are here. We have my great friend, the incomparable William Tong from Connecticut. And we have the original TJ, Tish James from the great state of New York. We have Carl Racine from the District of Columbia. And I wanna, I just wanna, I wanna thank Carl and his team at NAG for putting on really just a tremendously inspiring conference last week, the Presidential Initiative that was held in Washington, D.C., Carl's Initiative about Combating Hate. It was substantive, and Carl, I don't say this often, it was inspiring. So we thank you for your great work and leading us on this important issue. Thank you, Carl. I know she may be here later. I think she is virtual this morning from Massachusetts. General Healy, more of Healy, thanks for coming. All the way our Eastern neighbor, my closest neighbor, General Camacho from Guam is here. And I wanna introduce and give a warm welcome, a Vermont welcome and a NAG welcome to the new kid on the block, New Hampshire's Attorney General, John Formella. John, where are you, please stand up. We welcome you to the family. We get a great conference. These are incredibly important issues. We have wonderful panelists, and even better, we've dialed up, I think just incredible weather over these next two days. You are on the shores of beautiful Lake Champlain. Two blocks east, headed up the hill. I think I said many of you that way last night is Church Street with a lot of shops, restaurants, bars. Please go up there. One block down, of course, is the waterfront with more restaurants and shops. There's a great bike path that I know many of you went for a run or a walk on. If you head a little bit south in Burlington, this what we call the Pine Street Quarter, you're gonna see more restaurants and more breweries. So please get out and explore Burlington and the surrounding areas today. I know I was talking to a few of you. It is probably peak foliage season right now in Vermont. And please ask many of us what roads to travel. And the road less traveled in Vermont often is the one where you're gonna find a traffic jam this week in Vermont. But if you wanna be stunned, if you wanna be inspired, get out in the road and drive throughout Vermont. There's a fellow in here who can probably give you the best directions. And that's former Attorney General Bill Sorrell, who's here, Bill. Thanks for being here with us this morning. So in addition to this conference and the conversations we'll have this morning, NAG has put together a wonderful reception for tonight at Hotel Vermont, which is just around the corner. I think it's gonna be outside on the patio. Listen, there may not be a better night to be in Vermont. It's gonna be warm. It's gonna be sunny. You may have some leaves dropping down. It's gonna get a little chilly at night. We may have a fire pit. We're gonna have good food. Maybe have something warm in your glass. As my father would say, it's gonna be a nice night for an evening. So please join us here tonight. It's incredibly incredible for me to host this conference here in Burlington. I grew up in Burlington and I grew up about six blocks down the road from here. And Burlington was a different place back then, Bill. And I don't wanna age myself, but it was a different place than what it is today. And we've certainly evolved and progressed in so many ways in this city, in this state, in this country. But growing up down the road in the south end of Burlington, I used to work at this corner store. That doesn't exist anymore. On the corner of St. Paul Street and Howard Street called Lodge Brothers Meat Market. And these places don't exist much anymore. It really was this community gathering place where everybody would come, particularly on a Sunday morning. It was a meeting spot. You had the very wealthy, you had the very poor, you had the middle class, you had the working class, you had everybody who came. And I was lucky enough to work there as a young guy. I worked for a fellow by the name of Mr. V. Mr. V had two rules. Customers always write, treat everybody with respect. And Carl probably was, and Tish was probably the best political education I've ever received because you interacted with people from all walks of life and you began to understand people. And to just, if you could talk to folks, and some folks like to talk a little bit more than others, but it was a wonderful place to work. And Mr. V had a system. He didn't take any credit cards. He had a system of credit where you would run credit on the back of a cigarette carton, interest-free. Never sent out a bill. People paid what they could when they could, but they always paid. And folks would come in as a young guy at 16, 17 years. I didn't quite understand it, but folks would come into the store, get some food for dinner, and you'd run the credit on the back of a cigarette carton. Never billed no credit, no computer. Trust. Man were a far way away from the corner of St. Paul Street and Howard. Because I was thinking about Mr. V this morning. Would he have insurance for a cyber attack? Would he know what to do if he had a notify or who to notify if there was a data breach? How would he protect his customer's privacy? Would he even know or be engaged that perhaps that information, that data that he was collecting was being bought and sold? And did he have an obligation to tell his consumers that? Did he have to register and understand whether or not he would qualify as a data broker under Vermont law? What would it mean for a customer of his to know that whether or not they could opt in or perhaps opt out? Of this system of big data. The world has changed rapidly. Technology has outpaced the law. No question about it. And the question that we're gonna confront in this room is simply this. Can the law keep up? Can the role of the attorney general that chief law enforcement officer charge with protecting consumers, charge with protecting children, charge with making sure that the marketplace is a fair and equitable marketplace engage in this new economy that is moving so rapidly where folks like Mr. V, his shop doesn't exist anymore. But how do we protect that consumer who may now be doing their shopping online, who may not ever qualify for creditworthiness, who may not understand to be sophisticated to enter into a marketplace of cryptocurrency and to understand what that is and whether or not they could be at risk of being scammed. We're also gonna explore what does it mean? What does privacy actually mean in this digital age? You know, you're in a state that, and I know General Sorrell would agree with this, that this notion of privacy, it's part of our DNA in Vermont. It's who we are. It's embedded in our culture. In fact, it's embedded in our state constitution. And trying to draw and define those rights is becoming increasingly difficult. We know technology is a good thing. We know that the only constant is change and that for a small rural state like Vermont in order to compete in a global economy, we have to be nimble. We have to be flexible. We have to be willing to engage in this new economy as a government. But how do you take those pillars of consumer protection? How do you take those pillars of making sure that we're protecting kids and infuse them in this new legal framework as we continue to navigate a 21st century economy? I don't have all the answers. I know many folks in the room do. So we look forward to a robust conversation, a collaborative conversation as we seek to find answers on behalf of consumers, on behalf of kids as we continue these next two days in this conversation. And I just wanna close by saying that it really is that incredible honor and privilege to be with my colleagues like William Tish and Carl and Levin and others in the AG community. And it's quite amazing for somebody who grew up in the South End to be in a room like this with you all today, to be down in Washington with Carl and folks last week to go through the halls of power of Congress and at times be able to visit the White House. But my worldview, I still look through that lens of Lounge Brothers Market of folks coming in and perhaps didn't have the money at the time to feed their family who worked all day, who were coming home just looking to take care of their family and to go to a place where they are not judged, but were accepted, were respected and were helped. That's my worldview. That's what I, that's the worldview I've tried to bring to the Attorney General's office. That's the worldview I bring to this conversation with you all today. I look forward to this. Thank you for being in Burlington. Let's have a great conference. Okay, well, listen, we're already behind schedule. That's the bad news. Let me, hey, listen, you're in Vermont. We do things a little bit different here, as you know. So we have a New York Times reporter, Kashmir Hill, who was with us this morning. But here's the thing. She's not gonna interview us. I've asked Carl Racine to interview Ms. Hill about these issues and her crowd breaking work on these issues of privacy. And so with that, let me ask General Racine and Ms. Hill to take the stage. Oh, let's get started. Yeah, TJ's with me. Oh, is you, how much? Ha ha ha ha ha ha. Ha ha ha ha ha. Settled here. Good morning, everybody. All right, really great to see everyone there. And let me just say, I don't know if there is a, you know, more authentic, decent, deep, intelligent, loving, caring, and you know, just overwhelmingly decent human being who I've met in my nearly seven years as Attorney General in this entire AG room than TJ Donovan. He is a truth speaker. He's a sensitive, caring, loving human being. And I think he's the right leader for so many opportunities, including his current job as Attorney General. So TJ, thank you so much for hosting us. And to the room, it's really nice to be back in physical presence with everyone and certainly appreciate everybody exercising their own personal choice with respect to masking inside and even outside or not. We all have to do what we need to do in order to keep our family safe. And we should be respected for those choices. I do want to take maybe five seconds to ask my team to stand up. My Chief Deputy, Jason Downs. That's really important to... Yeah, it's important to know these folks because they really do all the work and they perform at such a high level and they've been denied in light of the pandemic the opportunity to engage and interact and get to know many of you. And Jason, I think, is just an extraordinary lawyer. I'm not gonna get into any fights here about being the best Chief Deputy in the country, but I just said it. Emily Gunston is a great lawyer in my office. She's a senior counsel. And we travel large in DC so we also brought our Consumer Protection Privacy Ace, Lindsay Marks. There you go. Excellent. So look, this is gonna be about 42 minutes of a conversation around really important privacy issues. And it is a little odd, right? An elected official having a discussion, not really an interview with the New York Times reporter, but you will know if you don't know now at the end of the session why we are lucky to have Kazmir Hill with us, Kazmir Hill with us. She has been a reporter and focusing on tech and privacy issues since about 2009. She's currently, of course, at the New York Times, but has written previously for Forbes and a whole bunch of other publications. I think my favorite was your blog. What was the name of your blog back in 2009? The Not-So-Private Parts. Hey, hey. I like puns. Some people would put an R after that title, but all good. I have to say that her writing around the subject that we're gonna talk about, her writings in research and investigation, are, in my view, the most powerful, most incisive, most accurate reporting in regards to both the benefits and the dangers of facial recognition technology. Let me ask everyone here whether they open their smartphone by use of their facial recognition. Okay, so, ma'am, 40, 50% of the room is used to, in some form, facial recognition technology. I should also point out, and to be really fair here, that we're likely to talk about a company in particular, Clearview AI, artificial intelligence. And the reason why we're gonna talk about Clearview is not to bully Clearview or jump Clearview. It's for us to learn about what seems to be the most powerful software and artificial intelligence application around facial technology. And of course, being big, being powerful means you're gonna get scrutiny. And so that's why we'll be talking about Clearview. So, what, about 165 years ago, Justice Brandeis, who was a noted privacy scholar, even at Harvard Law in 1890, wrote a law review article about privacy. He'd later get on the Supreme Court, and of course, in the Olmsted case, declared words that I think many of us agree to, which is freedom actually means the right to be left alone. How many people agree with that? Well, enter technology, right? And here we go with our conversation now. Why don't we start off with this concept of facial technology? Is it a new concept? I mean, facial recognition technology has been around for decades now. They first started trying to develop this automated way of recognizing faces back in 1960s. Police departments have been using facial recognition technology for two decades now. It started in Florida and Pinellas County back in 2000. So, it goes far back, but facial recognition technology got really good in the last decade, thanks to advances in neural net technology. And so, it's working better than it ever did before, right now. And there's a couple of different kinds. There's verification, opening your phone, and then there's this identification of going through a big database and looking for someone. And the reason that Clearview AI has gotten so much attention is that they went and they scraped billions of photos from the public web. They now have over 10 billion photos of people on social media sites, newspaper photos, and created this app that can identify just about anyone. And that was quite a dramatic development because before facial recognition had just been applied, at least in the government use, to people who were in government databases, mugshot photos, driver's license photos. And so, that was this big change that's now, made people look more closely at facial recognition and where we are today and where we go from here. You mentioned the number, 10 billion images. I think there are about 7.8 billion people in the world. So, you're not saying that every single human being in the world has been captured by some kind of photo, but what you're saying is that there are 10 billion photos certainly of everyone here out there in the database at Clearview. And it's just a matter of time, frankly, before there are more. Is that fair? That is fair. I mean, I would encourage, I don't know, I don't know what the relationship is between different law enforcement agencies. But certainly, you know, in your states, there are police departments that have Clearview. I mean, if you can go to your local agency and ask them to run a Clearview search on you, I think that you should do it because it's very powerful to see it in action. To see, you know, not just identifying you, it's not just about putting a name to a face. It's about showing you all these photos of yourself on the internet that you might not know are there. Photos on Flickr, you know, photos maybe somebody scanned onto the internet from your law school days. What is so powerful about this kind of facial recognition is not just figuring out who someone is, but being able to associate these online dossiers that have been compiled on us now for decades, associating that with your identity. So all the photos of you that are on the internet that don't have your name on them, but you can find now because you're looking by face. And then all the other things that just come with that, you know, the rating systems, whether you're in debt or not, these things will start to be able to be associated with your face as you move through society if we decide to make this more universally available. Indeed, in your article, your face is not your own. You highlight the common uses of technology like clear views for law enforcement. And you talk about cases like child exploitation cases. Have the technology been successful in these cases and has the technology assisted in the closing of criminal cases? Yeah, I've talked to a lot of law enforcement officers who have used facial recognition tools, and they have, I mean, there are some incredible use cases. I wrote in the article about a, there was child sexual abuse material where they had this photo of a young child being abused and the man's face was kind of briefly visible in the material. And, you know, you can't often get a lead from something like that, but because of clear view, they were able to run the man's face and they found one photo of him online, an Instagram photo that wasn't of him. He was in the background of the photo. It was taken at, it was taken at like a, what's that? I think it was a workout facility. Yeah, it was at a bodybuilding conference. And so then the officer, you know, what officers always tell me is that facial recognition isn't going to close the case. It's, you know, it's a lead. And so they contacted the bodybuilding, you know, vendor that this guy's standing behind the table and eventually they're able to get this guy's name and find him. And that's the thing about facial recognition. It can be very powerful. It really can help solve crimes, but it has flaws too. And I just don't think that we have fully thought through, fully thought through those flaws and how it will impact society. Bias especially, there just have not been enough studies done about how facial recognition works differently for different demographics. Let's talk about that in the criminal context. We've also seen wrongful arrests, wrongful convictions. I think in the literature, including your own reporting, there are at least a handful of such cases. They tend to concern people of color, Asian, Pacific Islander and black and brown. Has that at all served as a check and balance governmentally in regards to the use of this in law enforcement? You know, we, How accurate are they? How accurate are they? So there really has not been enough study of how accurate they are. The best study was done by NIST, which I hadn't really heard of before I started doing facial recognition studies, but NIST is the scientific lab, the federal agency, and they've been testing facial recognition algorithms dating back to the 90s. And for the first time in 2019, they did this extensive study about how demographics affect facial recognition algorithms. And they found that with one to one, with verification, there were problems with people of color, Asian people, black people, the elderly, the young, that it didn't perform as well on all groups. And with identification, where you're searching a big database, it didn't work as well on black women. But the thing is, they were just studying mugshots. They were studying mugshot to mugshot, driver's license photo to driver's license photo. And when we're talking about using facial recognition and criminal investigations, we're talking about an ATM photo, a surveillance camera still. These photos are not as good, not as perfect as a mugshot. And we haven't really studied that fully, how bias plays a role there. So I don't think we really know how bias there are, but the studies that have been done indicate that there are problems. And certainly those folks who wrongly arrested and wrongly convicted think there are significant problems. Yeah, I don't know of a wrongful conviction yet. There is somebody in Florida who alleges that it was a bad facial recognition match, and he's been trying to fight it for a long time. But I have talked to two black men who were wrongfully arrested based on a bad facial recognition match. And we don't know how many more there are. There's at least one more. Strangely, they're all in Detroit. In Detroit, you're asking about what law enforcement has kind of done in reaction to this. And in Detroit, they put some restrictions on how they're using facial recognition, that they're only using it for serious crimes, and really making sure that it is being treated as a tool that assists in the investigation and isn't the end all be all in it. And some cities have banned facial recognition for now. They say the problem of bias is too big a problem for us to be using this in the criminal justice system right now, and so around Boston, there have been bans in the San Francisco area. And they say, basically, yes, we see the usefulness of this tool, but we need to fully understand the bias problems first and think about the implications of this new technology that is changing our notions of privacy. Indeed, in Vermont, of course, also passed legislation a few years back, banning the technology, but more recently, there was discussion around exceptions allowing for child exploitation cases, seeing some value there. So the discussion in a real way is just beginning. Clearview AI, kind of a catchy name. Was that its name originally? No, when the company first started, they were called Smart Checker. Smart Checker. Without the, you know, in the Tech Valley way, without the E on the ER. That's how my niece writes it. Yeah. All right, let's move a little bit to the social media companies. You mentioned that the way that Clearview gets the pictures, the 10 billion plus images they scrape, I guess, that's what it's called. Social media and open internet sites. How have the big technology companies themselves, masters of AI reacted to what some would say their products being utilized in this facial recognition manner? So they have expressed displeasure about it. Facebook, Google, LinkedIn, Venmo, which are sites we know have been scraped by Clearview AI. They've sent the company cease and desist letters, but that has been the extent of it so far. You know, these companies, you know, this is hard to protect against. These companies do not like scrapers generally. They tend to have technologies that prevent that. They do something called rate limiting where you can basically see that the same computer is coming to your site over and over again and basically downloading photos, copying and pasting information, and you try to block that activity. But scrapers have, you know, it's kind of a cat and mouse game and they find ways around that. So it is hard for them to perfectly protect users' data. Sure, and some of the companies that you mentioned, Google, Facebook, of course, they've got great technology themselves. Is there any information or suggestion or evidence out there that they themselves have considered entering into the facial recognition technology space in a way that Clearview AI has? I mean, these companies already do facial recognition, right? Like what Facebook, Facebook back in 2010 released something called, what was it called? Phototagging or Face Tagging, where they would tell you, oh, this is your mother, do you want to tag her? This is your friend, John, do you want to tag him? They had these technologies already. And I'm sure that a Google or a Facebook could have released a Clearview AI type tool years earlier. Google actually said back in, I think it was 2011, the chairman, Eric Schmidt, said, this is the one technology that we have developed and we decide to hold back because we think it could be used in a very bad way and in a very good way. So it is something that the technology companies could have done themselves and they basically saw us taboo, which is pretty wild. I mean, when you think of Google and Facebook, you don't think of them as super conservative when it comes to use of people's data in privacy. So I think that that is really telling. And on scraping, I was just going to point to the fact that there's been litigation over this. There was a company called HiQ that was scraping LinkedIn and they basically sold the service to companies where they could say, hey, we can let you know when your employees are thinking about leaving. And they did that by basically scraping LinkedIn and looking at who had the fact that they're looking for a new job and stuff like that. And so LinkedIn sued them and said, you can't scrape us. It's violation of our protective material. And that case went to the Ninth Circuit and LinkedIn lost. The Ninth Circuit found that this is kind of publicly available material and there is a right to scrape. And so that decision actually came down while Clearview or Smart Checker was in the midst of scraping people's photos. And so they were pretty happy about that, that it said what they were doing was not legally wrong. And that is something that's kind of still undetermined, the legality of scraping among the many other things that are undetermined here. Right. TJ, the second over there, the attorney general from Vermont, mentioned that the law is behind the technology. I know you're not a lawyer, but you certainly know law. There's a 1986 fraud and computer access law. What's your sense as to whether that applies to federal law? I mean, that is part of what LinkedIn tried to use in this IQ case and that failed. That I think just because of the, this is information that's displayed publicly. And Clearview always makes that distinction. They said, you know, we're not logging into Facebook and getting access to private photos. We're not scraping places that anyone can't see. We are just like Google. We are like Google crawling the web and Google goes out and goes to everybody else's websites and gets their images and so that you can do a Google image search and you can look for general Carl Racine and see all the images of yourself that are on lots of different sites. Clearview says, that's exactly what we're doing. We're doing what Google does, except instead of searching for Carl Racine, you can take a photo of yourself and see all the images out there on different sites. And so that is, you know, that's a powerful, that's a powerful argument, I think. Yeah, and so obviously we're talking about federal legislation, that's old, 1986. What is the state of federal legislation that would tackle the issues related to facial recognition today? I mean, I keep seeing bills kind of get floated around facial recognition. Some people calling for a moratorium, for studying it. Some people calling for a ban. Some people kind of, some people looking to, I mean, so Illinois is the state to look to for a bill, for a law that affects facial recognition because they have the Biometric Information Privacy Act and this is quite a distinct law. Two other states have kind of a version of it, but it's the strongest in Illinois because of the right to private action. That's right. And we have the former Illinois Attorney General with us, our great friend Lisa Madigan. So we may have a question from Lisa later. The Illinois law is particularly interesting because it's an old law. It's a 2008 law. 2008 and it gives people the power to decide whether their biometrics get used or not by private companies. Consent. Affirmative consent. And so this has really been the biggest problem for clear view AI is the existence of this law and that's where they're being sued. There's a bunch of class actions that have been consolidated in Illinois. There's a state lawsuit filed by the ACLU and that's what they're fighting is this law, the fact that they didn't get consent from Illinois residents. They've now gotten Illinois residents or they've tried to get them out of their database. They're not operating in Illinois, but I think that's the big question is whether BIPA is the way to address face recognition or not. Right, but from what you know and what you've reported the Illinois law has caught the company's attention and at least they're trying as best they can or at least that's what they say to not have residents of Illinois' images in their database. So it could be a starting point for other states that are similarly concerned in regards to the use of facial recognition without people's consent. And in California of course with their big privacy law, people who live in California can contact ClearView AI and have their photos deleted from their database. Interesting. So that's another law that applies. Okay, so we've been talking a lot about ClearView AI and facial recognition technology generally in the government space and specifically criminal space. I do wanna move over to the potential corporate space and commercialization of it but before we go there, facial recognition technology is also powerful overseas and being utilized by countries like Russia and China. Can you talk a bit about how those countries are using AI, specifically facial recognition AI? Yeah, I haven't, so this part I haven't directly reported on as much so I have to rely on other people's reporting but China and Russia are pretty far ahead of us in terms of the deployment of facial recognition. Russia for example, they had a kind of ClearView AI type app come out back in 2016, it was called FindFace and it was publicly available and it was based on, they had scraped a social networking site called VK which is like Facebook and people could download the app and start using it and so people went on the subway, they identified strangers, harassers took the faces of porn stars, sex workers, identified them, started harassing them and their families. It was used to identify people at anti-government protests and the company eventually took the app down and didn't make it publicly available but they did sell it to governments and to businesses and now their technology is deployed in surveillance cameras around different Russian cities and so they can do real-time facial recognition, they can find people who there's a warrant out for their arrest or it has been reported that it was used during the pandemic that people that were supposed to be quarantining if they left their apartment, it would basically flag that they were outside when they were not supposed to be and one guy said he took out his trash and then the police came to his door and asked why he'd been out of his apartment while he was supposed to be quarantining and so when I look at what's happening in Russia and China, I mean it seems like a possible future depending on what we decide about what we do with facial recognition in the US but yes it's quite advanced in its deployment there already. And indeed in China there's reporting out there that the Chinese authorities have been using the technology in regards to the persecution of the Uighurs that's the Muslim followers in China and so yes there is a real concern around government misuse albeit more specifically abroad though problems have been noted. And in China they're doing something, facial recognition but there's other kinds of facial technologies and so one thing that was reportedly being done in China is that they've developed face recognition that can basically identify whether you are a Uighur Muslim and so there can be an alert that a person of this ethnicity has entered the premises and so that's troubling in its own way. To say the least indeed. Let's talk a little bit about corporate use and worst case scenarios and dig in a little bit more about Clearview AI. Who founded the technology that is now Clearview AI? So Clearview AI was founded, I've gotten kind of different stories on this from the company originally my understanding is that it was founded by a man named Richard Schwartz who was kind of a political media consultant he had been a deputy to Rudy Giuliani when he was mayor of New York and a technologist named Juan Tan Tat who was from Australia had moved to the US when he was 19 to work in California and Silicon Valley making Facebook games and apps and then he moved to New York and they met and they developed the company that became Clearview AI. Later I found out there was a third person that was involved in the founding named Charles Johnson or Chuck Johnson who is a very controversial figure kind of associated with the alt-right and that when it was first kind of conceived he told me it was about kind of identifying liberals that it would be they all met he and Tan Tat had met because of being in those circles they both really liked Trump they like went to the RNC convention and were thinking this would be a good way to be able to identify liberals who were kind of like sneaking in here and so that is, it was a surprising kind of origin story and the company says that this isn't there seems to be like quite a disagreement between the company and Chuck Johnson now though he still does have some equity in the company according to corporate filings that was able to track down. And who's Peter Thiel? Peter Thiel was the first investor in what was then called Smart Checker and became Clearview AI. It was a very small amount for a billionaire like Peter Thiel was like what you might spend on a latte he gave them $200,000. I'll take a latte. But Peter Thiel you know he started PayPal he started PayPal he made a fortune investing early on in Facebook he helped found Palantir I mean he is a person that really understands the value of people's data and so it, yeah perhaps not surprising that he was a backer of a company like this. Indeed and you mentioned the political aspect I think I've read in your writings two points number one that this facial recognition technology has been shopped to political folks people running campaigns perhaps to be used in opposition research and the like. Tell us about that. Do we know whether campaigns are actually using facial recognition for campaign purposes? So in the early days of Smart Checker they didn't know what they were gonna do with the tool that they are creating and originally it was not just about face it was also about searching somebody's email but they thought about selling it to grocery stores they did a pilot in Cristetes market which is in New York they were selling it trying to sell it to real estate firms that you kind of have it in the lobby and identify people who are coming in and originally the company tried to pitch their technology to politicians that were campaigning one was a woman named Holly Lynch who was running as a Democrat in New York another was Paul Nealon who was running in I believe Wisconsin who later came out pretty explicitly as a kind of white supremacist or at least white supremacy friendly and the version that they were selling to them Paul Nealon never talked to me but Holly Lynch said it didn't have anything to do with facial recognition it was just about this kind of tool that would tell you more about voters based on their social media but you could, I mean certainly imagine very useful ways to use it for campaigns as far as I know there are no political uses of face recognition right now and certainly not of clear view AI but there's a lot in the world I don't know for sure I can tell you that I've had an interaction with clear view AI they sought I think a meeting I don't know if it was virtual Jason or not with us and they gave us an overview of their technology I believe Mr. Schwartz was present and I believe that Mr. Tott was also present and specifically what they really wanted to communicate was how their technology was being utilized to identify the mob at the insurrection on January 6th and again making that strong law enforcement pitch I'm sure that they've reached out to other AG in that regard as well Did they do a demo for you? I don't remember No but let's talk about a demo because you and your reporting working sources asked that your face be run on their database tell us about that Well when I first started reporting on clear view AI I'd heard about it they'd shown up in a public records search that a researcher had been doing about how police departments were using facial recognition and this company clear view AI showed up and they had a legal memo there giving the police departments I was written by Paul Clement the former US solicitor general explaining why this tool is actually legal to use that they wouldn't be breaking the law by searching someone's face using clear view AI and so I was like, oh this sounds like nothing I've heard of before I tried to get in touch with the company they wouldn't talk to me no one would respond to my emails they had an address on their website and when I went there it was a couple blocks away from the New York Times office the building didn't exist it was a very odd they were in hiding they say it was just because they were in stealth mode as a startup but it was very strange for me as a reporter I'd never quite had that experience before and so I started trying to find police departments that were using it so I could talk to police officers and see whether this thing actually worked and I talked to a financial crimes detective in Gainesville and he was very excited to talk about it he said he wished he could be their spokesman that it was amazing that he had dozens of cases from photos where he hadn't been able to get a lead checking state facial recognition databases and that he ran the photos through clear view AI and all of a sudden he was identifying these people and I said well you know I'd really like to see how this works myself and he says oh yeah I'd love to run a search of your face just send me a photo and so I did and then he stopped talking to me and I would email him I called him and I couldn't get him to respond so I was talking to another investigator in Texas who wasn't quite as big a fan of clear view AI but said it worked really well as long as you had somebody who was on the internet and I was like oh I talked to this other guy and he never got back to me but and he's like oh I'll run your face and so he runs it and he says oh there weren't any results and I was like that's so weird there's a lot of photos of me on the internet and he goes yeah it's really weird maybe like they're servers down or something and then he stopped talking to me so then I find a third guy and he I tell him about how this happened he goes okay well let me check and he checks and he says yeah you don't have any results it's so weird and then a couple of minutes later clear view AI calls him and says hey you know this is a clear view and the company calls the law enforcement officer calls this detective and they're like oh you know we have some questions why are you running the photo of a New York Times reporter and he's like oh I did and they go yeah Kashmir Hill do you know her and he's just like oh I don't how would I know how would I know he was a little bit cagey but there said you know you're not supposed to do that you know you shouldn't be talking to the media about clear view AI and it was stunning to me because this company wasn't talking to me but they had flagged my face in their app and they had blocked my face from having results and that I found very I just found that so alarming that you have this private company that you know kind of has access to everyone that law enforcement is looking for that's you know quite powerful but you know that's this kind of larger issue of the way private companies are assisting government agencies but yeah that was quite a strange start to the story of telling clear so we got about eight minutes and 20 seconds or so I wonder if you can just delineate or list some of the grave concerns around more than just the government use but corporations monetizing this app and using facial recognition technology yeah I mean I think it is when I think about facial recognition and the usefulness it could have the utility in society you know I could imagine this being released more widely to the world I mean you certainly are already seeing it in airports, airlines kind of letting people use their facial right their face to check in a lot of us use facial recognition all the time with our phones I can just see the way it might creep into society more broadly you know I imagine that we could all have an app like Clearview available to us at some point that we might have it on our phones that we would use it on each other and that would just completely change our sense of anonymity in the world that you if you're in a restaurant and you're having a juicy conversation and you're able to do that because you just assume that the people around you have no idea who you are with the Clearview app they could just run your face and now they understand the context of the conversation that I find yeah you know almost more you know one of the more more chilling ways that it could be used at the same time as a reporter I would love to have access to a tool like this like this is the thing about privacy we we you know vigorously want to protect our own but we love ways to invade other people's but like for a reporter if there's some kind of event and there's people standing around and you weren't there but now you can identify those people and go and interview them you know it's a useful power but it would be quite chilling to not be able to just walk down a city street to to think of the way I you know I've I've written before about people who who basically have interactions with strangers where they get very upset at them and they try to destroy their lives you know go online and destroy their online reputation and I just think about the way that these slights could blow up you know you're on the bus or you're on the subway with somebody you bump into somebody you have angry words and all of a sudden they can know who you are and you know write horrible things about you online I just imagine you know there are positive cases or negative cases every technology humans get we use it for good and we use it for evil and so indeed and you know you're not to mention of course the proliferation of social media and kids and you know the fact that kids like taking pictures of themselves and their friends you can imagine how that might be used in the wrong hands we got about five minutes or so left for audience questions let me first invite the attorneys general to chime in if they've got any questions or comments Tisha I'm gonna give you a mic so I was in the Berkshires recently enjoying a dinner by myself and watching a football game and it was an innocent interaction the gentleman to my right were having a private bet as to whether or not that was the Tisha James they did a facial recognition and then they came over to me with the phone and said you are the Tisha James I just bet my friend and I won $25 or whatever we did you know we took a picture of you and there you are and then they went on to do was ask me a series of questions it was scary was somewhat alarming it was innocent then but in the hands of individuals who may want to do you arm it could be somewhat threatening so my question really is racial recognition and the bias is associated with the algorithms and the fact that the accuracy rate they claim when we did meet with Clearview that the accuracy rate was over 90% they claim or about 90% but when you inject race into it gender, age the accuracy rate goes down and so as was mentioned by A.J. Racine very much concerned about the use of algorithms particularly in the criminal justice space and if you can speak a little bit more to that that would be greatly appreciated but again I personally experienced it and it was somewhat alarming in that incident innocent, it was harmless but clearly in the wrong hands obviously it poses some concerns yeah I mean the problem with the technology that has made facial recognition so good now is that it's neural net technology and neural net technology the technologists kind of talk about it as a black box technology because you have try to think of the simplest way to describe you just have like layers of computer programming that you feed a bunch of data into it and give it the right answers and it basically figures out how to perform face recognition because you give it a whole bunch of photos of the same person and it figures out how to analyze it and when something goes wrong with this technology it's hard for them to fix it because it's all this like layers of computer programming they can't just see where it went wrong and make sure that formula is right so once you get the bias baked in it can be very hard to take it out and we really haven't had enough studies done on this and a lot of technologists don't even understand where the bias really comes from is it the training data was it trained on a bunch of photos of white people and there just wasn't enough diversity in the data that they fed into the program or is it cameras themselves which have been optimized to take photos of white skin and not brown skin or black skin it is hard to kind of figure out where in this process the bias creeped in and so then it is hard to take it out I usually wanted to ask you about that generally seen because I I've heard you might have I've heard that you might have some kind of bill for requiring companies to figure this out to analyze their own algorithms for bias and I'm curious how that might work will it be a feature in the Sunday Times I've got my notebook no the fact is that the District of Columbia our terrific office has been working with experts particularly Georgetown University privacy experts that you're familiar with and civil rights groups and we are close to finalizing an algorithmic fairness and transparency act for the District of Columbia we're not gonna put it out there until we fully vetted it there are a lot of issues and concerns and of course you wanna be fair but I can tell you that we've gotten really good input from the corporate community and we think we're on pace to moving that forward and should be interesting to Attorney General James's point the algorithms themselves have problems there is no doubt about that they oftentimes rely on age old bias and they put it into some kind of computer language and act as if that's objective criteria it's not, it's based on human bias that has been going on for a long time so there are significant issues there the other point that Attorney General James raises is who's being surveyed the District of Columbia in the latest census data 2020 reported that the population of African Americans in the District of Columbia significantly dropped over the last 30 years down from about 82% in the District of Columbia to 41% estimated in 2020 I can tell you that in the District of Columbia in regards to juveniles arrested coming into my office 96% of the kids are black, brown kids I can also tell you it ain't because with all due respect to my white brothers and others it's not because they're not violating law it's because the government is focused on particular communities of color so algorithms are problematic as well as who the focus of surveillance is on I often joke about the times in Wall Street you know the big financial crisis of 7.08.09 boy what if they all of a sudden allowed for stopping Frisk of any banker with a briefcase under the theory that you stole money from grandmothers and grandfathers I'm sure there'd be an uproar at any rate these are fascinating issues I'm gonna take a Vermont privilege here and go a couple more minutes are there any other attorneys general who want to ask a question and I'll expand that to former AG and then we'll go out to the audience Thank you and thank you for being here today thank you general and thank you Tish for sharing your story and what I took from that story is in that instance though scary they got it right you are Tish James but for all of us and I'm sure you've had this experience who are people of color we are often routinely mistaken for other people right who are our same race and gender and I'm routinely mistaken for other Asian-American elected officials and there aren't that many of us to be very candid and it happens all the time and so this is very scary that even though that's human error that the biases built in could result in machine error I actually also had a meeting with Clearview it was a more casual meeting at an event and one of the founders came up to me with his phone and said I've got this great technology let me show it to you and we're literally like at an event cocktail party and then he wanted to give me a demo on the fly and that just was beyond creepy that he had this power in his hand and he wanted to demonstrate it on the fly in a cocktail party and that seemed to be literally in one person's hand a great deal of power, more power than frankly I had encountered in this space so with those observations my question to you is we've identified the problems I think we're here to try to figure out what do we do about this and so in your travels in talking to government officials, regulators, deep thinkers about a regulatory infrastructure rules laws is there something that you've come across a bill or an idea or a regulator, commentator who said something that resonates with you about how we approach this problem from our regulatory and law enforcement perspective and trying to keep people safe and protect their privacy? You know I haven't, I don't encounter a lot of people that have the answer because I do think it's so complicated. Right now there's kind of a knee-jerk reaction which is just ban it and some places have but it is, I think bans are difficult given that it can help to solve crimes that it might make it possible to solve a crime that you otherwise couldn't solve that that alone makes it difficult to do a full ban. A lot of people point to BIPA and BIPA is kind of useful for the regulation of private use of facial recognition. There's an exemption for government use in BIPA. With so many privacy problems that often comes down to consent and maybe this is fine as long as you opt in. What if there's some big facial recognition app network and we all enter it and you can have different settings so that your friends can recognize you or friends of friends can recognize you or anyone can recognize you. There just aren't often easy ways to solve privacy problems especially under the framework of existing laws which just didn't anticipate these kinds of technologies existing. So I haven't figured out the solution yet. I mean, I hope I do because I'm working on a book about this and I'd like to. Coming out soon, Random House publication. I'd love to end it with the answers but I do think it's just very difficult. I often, I've thought a lot about caller ID that when caller ID first came out some people thought of it as a privacy invasion that when you called somebody you would be identified. That we should have a kind of anonymity when we call people and how quickly that shifted. That idea now when you get a call from like an unknown number or a blocked number, it's annoying. You wanna know who's calling you and so I can, I think it's possible we could have that shift potentially around facial recognition where we have the expectation that you can recognize someone. I just think it's hard to anticipate how societal attitudes are going to change but I don't unfortunately don't have like a simple answer to that. It's complicated for sure and I think quite honestly we could have a four day conference on this issue. There's no doubt about it. So we're gonna end here. Really wanna thank Kashmir Hill for her extraordinary work and willingness to meet with this group. Thank you very much.