Who the Machine Serves

Hi everyone. Thank you so much for joining us for the latest segment of Affecting Change EFF's live stream series where our panelists tackle the latest issues in civil liberties and human rights in the digital world. Before proceeding, we have a few housekeeping items to go over. We have a live caption team online with us today. You can view the captions at liveream.eff.org. there. You can participate in the Twitch Twitch chat and if you have questions throughout the program, please drop them in the chat and our guests will answer as many as they can at the end of the stream. If you're comfortable, we encourage you to type a hello in the chat window now and let us know from which part of the world you're joining us from. For those of you who aren't familiar with the Electronic Frontier Foundation, EFF is a nonprofit composed of technologists, activists, and attorneys. Since 1990, EFF has been fighting for your right to create, innovate, and make yourself heard by anyone willing to listen without fear of unjust government or corporate surveillance. You are the reason EFNF can take on this important work. And if you'd like to help us in the fight to defend your digital freedoms, please consider donating at eff.org/supportchange or pick up some cool EFF merch at eff.org/shop. My name is Sarah Hamid. I'm the director of strategic campaigns at EFF where I lead our municipal legisl legislative work supporting lawmakers, community groups, and local agencies who are facing real decisions about whether and how to bring AI into their communities, especially in public safety contexts where AI enhancements are being integrated into existing surveillance infrastructure with little public scrutiny or oversight. It's my honor today to introduce the participants in today's discussion. Nicole Ozer became in June 2026 and is a legal expert on artificial intelligence, privacy, surveillance, and digital speech. Earlier, she created the Center for Constitutional Democracy at UC Law San Francisco, and before that was the founding director and longtime leader of the ACLU Northern California's Technology and Civil Liberties Program. Corey Doctor is a science fiction author and journalist, and he's been an activist at EFF for nearly 25 years. His latest book, The Reverse Centaur's Guide to Life After AI, is out now. As a reminder, we'll have about 15 minutes for your questions at the end. So, if we haven't touched on something you want to hear about hear about, please ask. And you can submit your questions via the Twitch chat at any time. So, before we dig in, a quick word on where EF F sits in all of this. AI touches nearly every issue we work on right now. We're fighting copyright lawsuits that threaten beneficial AI uses, pushing back on the FTC's attempt to police accuracy in AI systems, defending crawlers and data systems from misguided state bills, and scrutinizing how AI features get quietly baked into your phone without you having real control over what they can see. AI is showing up across our litigation, our policy work, our activism, our tech projects simultaneously. And that breadth is exactly what today's convers why today's conversation is going to move across a lot of ground. Labor, privacy, copyright, government use. AI is not just one thing. It's not a single technology. Um it's a label that we stap on everything from hiring algorithms to a chatbot to a license plate reader with a new process layer bolted on. And this matters because of how AI often gets more attention than the the how of AI rather gets often more attention than the who. who deployed it, who profits, and who gets a say in how it shows up in their workplace, their insurance rate, or their neighborhood. I see this play out constantly in my own work, where I spent most of my time engaging directly with city council and local decision makers who are whose buyin is being sought for an AI enhanced tool or workflow, often with just a comment period and a sales desk deck, let alone a risk assessment. And that's one of the through lines I really hope we pull out on today. Power over AI isn't just con contested in Washington or Sacramento. It's contested at a Tuesday night city council meeting in a school board's vote to buy a monitoring tool most parents never get to hear about until it's installed and in private purchasing decisions we that never go through any public process to begin with. So as we get into surveillance pricing, algorithmic bosses, and government AI and what good AI might look like, I'd ask everyone watching to keep asking the same question at every scale. Who does this serve and who gets to decide? And with that, I'll turn it over to Corey. Oh, thank you, Sarah, and and thank you, Nikki, for joining and and hello to everyone out there. My goodness, I'm told there's there's 900 of you, which is uh that's a that's a big crowd for a live stream. Hi, chat. Uh, so Sarah, I really liked your introduction because you hit on one of the themes that as a recovering science fiction writer, I find most important about EFF's work, which is that um, we are so much more concerned with who the gadget does things for and who it does it to than what the gadget actually does. And I know that's an area that uh, Nikki is focused on as well. So Nicole, you know that uh, AI is being used increasingly in corporate and government decisionmaking. uh what would it look like if that AI was you know uh organized around the needs of people rather than you know um streamlining uh corporate processes so that bosses don't have to have ego shattering confrontations with people who know how to do things and tell them their ideas are stupid and can be replaced with software and if you know you could have uh people running governments who wouldn't have to talk to the permanent civil service who told them that their ideas were stupid and illegal and would kill people and you could just ask the algorithm to do it. What would it be like if if if we had statistical inference tools that were deployed in a way that was organized around like the needs of the people who are being processed by those statistical inference tools. >> Yeah. I mean, hi everyone. It's so good to be here. Thanks Corey. Um was really excited for this conversation. I'm coming to you all from EFF's headquarters in San Francisco and you know I became executive director on June 1st. So just a couple months into my tenure but you know this is really the critical question of this moment right this is what EFF was really created for. It's about ensuring that technology supports rights and justice and freedom um and innovation and this is this consequential moment. So I come in as the new executive director. You know, Cindy has passed the torch to me to lead EFF. And this is such a consequential moment because we are at this moment obviously that there is, you know, many of the threats to rights and justice and freedom and democracy that were once thought to be hypothetical are the reality right now. And we're also at this moment where the decisions that we're making in the courts, in the legislature, in our communities, in our daily lives, um, about how AI is developed and how it's used is going to determine not just our future, but the future of, you know, the next generation. Um, and we come at this moment and sort of have this conversation today. uh when you know at base you know governments and large corporations possess surveillance capabilities that were just unimaginable certainly unimaginable when EFF was even created in 1990 unimaginable even when I started my career just post 911 when there already had been an exponential increase in government surveillance and actually unimaginable even just a few years ago um you know this trajectory of AI has been so quick. I mean, you know, Corey and I have been around a long time seeing sort of these increases in technology, but the rapid pace of how this is iterating, I think, has even been surprising not just to us, but to many of the companies that are that are building this technology. So, you know, I always talk about, you know, our rights are protected by law or they're protected by friction. And we see AI as as Corey said, you know, AI is really intersecting with so many parts of our lives. And the questions about, you know, how is it going to be used? How is it going to be deployed? How are we actually crafting the kinds of rules, regulations that really protect innovation, that protect the ways that AI can actually benefit people, but also think about the ways that we can make sure that this opportunity that this, you know, um this advances in technology can actually lift all boats rather than just profit for a few. And so those are things as I come in as executive director um and as Sarah said are deeply on our mind across strategy at EFF across issue area work we're doing in the courts work we're doing in the legislature work we're doing at the community level and how we can really make sure that AI really is working uh for the people and that we are developing and using it in ways that protect fundamental rights and work for those who build it uh use it and are affected by it. Um and you know my uh you know I have been in public interest technology my entire career and I believe deeply in um you know working together figuring out ways uh that we can build power across issue and strategy and honestly we are at this pivotal moment right now like the future of AI hasn't been written and we can actually work together to get it right. >> Yeah. You know, I would add to that because maybe, you know, I I think we we have uh we might we might feel slightly differently about AI and that I I think that it's a really interesting technology that will like we're going to be using advanced statistical inference tools for as ever because they're useful. I don't think AI is going to change everything. But here's what I worry about. I worry that the fight about AI is going to change everything. I worry that people who want to know what NBC looked like before it was stolen and given to a crony of the president are not going to be able to know because we've banned scraping in the name of ending AI. I'm worried that we are going to um institute all kinds of restrictions about image generation uh in the name of preventing defakes. uh that will make it harder for people to create parodies and will create pretexts for the removal of content that makes powerful people uncomfortable. Um you know I I I like like you're saying uh we've both done this for a long time. I mean I also started right after 911 and after a quarter of a century you kind of wish that you lived in a society that had enough object permanence to win a game of peekaboo and then they could remember the last time we were like you know doing this security syllogism. something must be done there. I've done something now something has been done rather than making it something that actually addresses your problem that we all ended up really >> I know I know what's to me it's it's what's old is new again uh a lot of sort of what we're seeing in the conversations we're having and I I think that's why um you know as Sarah said you know EFF is working across so many different issues and I think that's really unique because as you said Corey there are folks who are kind of looking at one area of AI or another area of AI and they're they're kind of weighing things just for that issue. But EFF is not only brings together and this is sort of the superpower of EFF is that, you know, it's it's lawyers, it's activists, it's technologists, it's community work. Um, and it's not just, you know, it's privacy, it's consumer privacy, it's government surveillance, it's creativity and innovation, it's intellectual property, um, it's competition. And so we are actually looking across these areas and not sort of putting the you know our weight on one side without thinking about what pops up on the other side. Um and you know we know a lot of folks in the field and and there are folks you know who are thinking about one issue without maybe the ramifications on the other side of the issue. And and that honestly does take both sort of that deep expertise but also that historical memory of like I've seen this before, you know, uh I saw this in early 2000s. Let's actually learn from that history. Let's think about what did we get right? What did we not get right? What were we able to push? What we what weren't we successful in doing where are the stakes even higher now to get it right this time and to build the power to get it right this time? because a lot of times we actually there is an answer but we haven't had the power to get that answer through the courts or through the legislature. Um, and that's why I think the stakes are so high right now to like to to make sure that we can we can build it in the right way and get in those protections here at the ground level rather than kind of um, you know, it again being a technology that has the potential to lift all boats but instead ends up perpetuating sort of power in the already powerful. So, one area that we've really spent a lot of time working in that bears directly on this is uh public officials who get suckered into buying magic beans uh from from Gabby technologists. Whether that's uh internet connected voting machines or uh backdoor encryption that's somehow still secure or software that can automatically detect copyright infringement. all of these things that are like just you know uh pipe dreams and nightmares that don't work and can't work and that nevertheless we see a lot of public money being spent on and if that wasn't bad enough we then see it being pressed into service in in publicly important applications. So this is a thing we see with AI and I know one of the areas you've been researching and working in is is AI procurement. This process by which we wake up one day and find out that the uh town council uh has been taken for a ride uh and been sold some algorithms, some flock cameras, some automatic guilt uh generators and something to uh spy on all the kids at school. Um so where are you at and and what is EFF doing on this procurement side? What would a good procurement process look like? Yeah, I mean we have seen this you know again what's old is new again sort of every single cycle really since you know post 911 you know huge amounts of you know government funding coming down uh you know $1 billion dollar a year since 911 has come down from the federal government for quote unquote surveillance infrastructure um and a you know most of it has been snake oil not only just snake snake oil, but actually snake oil that harms communities, that actually makes, you know, makes people far more unsafe. that now we see those systems being weaponized against communities being used by ICE for, you know, uh, immigration, you know, attacks on immigrant communities, being weaponized against reproductive freedom, being weaponized against, um, uh, activists and, you know, who are speaking out against sort of the current administration and and we see sort of now a new cycle with AI. um you know most of these systems that have been deployed for surveillance have been done extremely quietly. The money was already in hand. There hasn't been engagement with diverse communities and real decisionmaking by you know diverse communities about the why and whether um and we see that right now with AI. We see the the jumping to the how of deployment of AI, the using of public money to purchase systems without the questions about why are we doing this, what are actually the risks, whether we should do this, any kind of evidence-based evaluation of what the benefits are and also what the extreme harms are. You know, we see there was an example, I won't say what city, but there was an example of a city I saw recently that rather than having a sign-in sheet at their community centers, they were going to use AI systems to record who came into their community centers. Okay. In a moment where this is a community that has a a very high immigrant population that then that of course that information is then collected and who knows where it ends up being used. Um, one one thing that's sort of a prime example of sort of how EFF looks at these issues across strategy is here's all this information now being collected on the local level about who people are and where they're going. They might be going into that community center actually to access resources on protecting themselves or signing up their kids for school or or health services. And the information that's collected on the local level uh here in California has then never been protected by the state law that's like the federal privacy act. It's called the information privac uh IPA. Uh it was passed in 1977 and that law has never been updated even though it it only applies to state agency databases. It currently doesn't apply to community, local, city, county databases. And so that's one of the things like we are looking holistically at in this sort of AI moment. We need to have real centering of community in the why and whether to use AI before the how. But we also need to think about the laws that are in place to protect whatever kind of information is collected. Make sure it's then not flowing uh to companies and to the federal government and being weaponized against against people in our community. So it's it's looking in sort of how we can address those issues also in the courts. Uh we are uh currently have a case using uh a constitutional right to privacy against license plate readers uh as well. And so we're really thinking holistically how do we use current levers? How do we really make sure that we are using both existing law butressing uh existing law and building new law and policy uh that will really protect people in the AI age? Yeah, I I love hearing about the salience of these local issues, not because it's great that we see towns with flot cameras proliferating and so on, but you know, one of the things we've learned from like the last five, six years of the culture wars is that some of the most unhinged weirdos you've ever met can make gigantic changes to their local communities, mostly for the worse in the last five or six years by joining poorly regarded local government agencies and standing for local offices. And you know, one of the, you know, top queries in my inbox from EFF supporters over the years is, "What else can I do?" You know, I've joined EFF. I I pay my subs. I get your mailing list. I fill in everything. And I think there's we have a a new generation of activists a morning. And you know, it turns out that those local offices are the difference between like whether ice chs get information about your neighbors, whether there's a flock camera outside your house, whether you know, all the portable water and energy available to your neighborhood is being diverted to a data center. Those are all uh extremely local issues. Now you you mentioned privacy and you know this is obviously another one of our long-standing campaigns and I I'll quibble a little with what something you said earlier which is that no one could have predicted even even [laughter] predict how bad this was going to get well I got to say I mean look this is this has been you know like credit where it's due. EFF's founders I think were really clued in to just how dystopian it could get. you know, you don't you don't found or work for an or like EFF, as you know, because you think everything's just automatically going to be fine, right? On the one hand, you have to be very excited about the potential, but you also have to be scared as hell about how it could go wrong. And I think the surveillance nightmare we're living through now >> is something that has been uh really uh uh something we've been warning people about for decades. Sometimes to the extent that people called us alarmist about it, but I think we're sadly quite vindicated. And as you say, you know, our privacy laws have not kept up. You know, California hasn't updated since 77. Federally, we haven't had a new privacy law uh for consumers since 1988 when Ronald Reagan made it illegal for video stores to store clerks to disclose your VHS rental histories, right? And that's not because we don't know how to write a privacy law. We we've written them before. We write them at the state level. Other countries write them. No one's asking our federal lawmakers to recover, you know, the lost practice of a fallen civilization. and they don't have to imbalm a pharaoh, right? They just have to write a law. And so, you know, that vacuum has created a lot of problems. And some of them have to do with AI. And you know what I find exciting about this moment of of people really waking up to the privacy consequences of AI is that it holds the potential for a coalition that isn't just people who are worried about deep fake porn, although that's a thing to worry about or being denied a loan or or a mortgage or a job because of surveillance data about you or uh having surveillance data surface the fact that you went to a protest or having surveillance data in used in ways that make you that make you feel that you know your loved ones have been brainwashed and grandpa's a queue on your kids anorexic and all the the the millennials you know are quoting Osama bin Laden on TikTok. I I don't even think all those things are real. But if everyone wants to fight for a privacy law because they're worried about them. I I'll be right there alongside of you on them. And one area where we're seeing some some people join the privacy fight is over surveillance pricing in California. And I know this is an area EFF has supported. Maybe I'll explain a little about surveillance pricing and you can talk about this this uh this bill AB2564. So, one of the things AI is is unambiguously good at, you know, there are a lot of things that that that we we claim AI can do, but one thing it's really good at doing is automatic experimentation and multivariate analysis where you take all the surveillance data you have on everyone who's going to buy something or everyone is going to sell their labor to Uber or a gig platform and you ask the algorithm to just bucket those people based on their behavioral data and demographic characteristics. You don't have to do what Don Draper did in Mad Man and like have a chart where you're like, "Oh, we're going to do moms over 50 in one segment or we're going to do depression haunted can stackers over here and high school dropouts who love shop class over here. We're going to try and figure out how to sell to anyone." You just ask the algorithm, find people who are similar. And then you just start automatically experimenting with seeing who will pay high prices and when. And you don't have to have a theory for this. You can just walk your way into the knowledge or the the the outcome that parents before 8 in the morning will pay more because they're trying to get their kids out the door. and that, you know, this is a real thing that our our friends uh at at Groundwork Collaborative and uh the Roosevelt Institute of Surface that nursing apps are taking nurses who have more credit card data and uh more credit card uh debt and offering them lower wages on the grounds that they uh will accept a a lower wage if they're in more debt. And and this is a thing AI is superb at. It can continuously test the lower bound of the wage you're willing to accept or the price you're willing to pay. And there are large commercial interests selling this into extremely concentrated sectors uh both um selling us goods and buying our labor. And the outcome is we're earning less and we're paying more. And so talk about AB2564. >> Yeah. I mean, so I think just sort of bigger picture, obviously sort of talking about, you know, our rights are protected by friction or they're protected by law. Um, you know, since the early 2000s, there has been this rise of surveillance capitalism. There's been a massive amount of information collected about who we are and where we go and what we [clears throat] do and who we know. you know, both by the physical surveillance that has been funded by the federal government, but also of course all of the electronic surveillance that we, you know, are now living digital lives. So, you know, in in 1990, um, most of us were not, you know, we weren't carrying a cell phone every day that, you know, said where who we were and where we were going. Um, you know, we were, you know, I remember, you know, you could still pay for your groceries with a check, right? Uh, or we were using cash. You know, it wasn't sort of a digital lives where we're showing up with our phones and, you know, paying for our groceries as we walked out the door. So you know we have this entire sort of surveillance capitalism ecosystem um which has been used in sort of the online space obviously to you know target advertising and create massive profit on sort of the big tech side but now we now have AI systems which are taking a lot of the things that are happening in the physical world that we're more protected by friction and we've got a conflation now and the ability ability to now use what we're doing in the physical world to also do the kind of surveillance pricing that we often saw already um in the digital world. Um so we are seeing things like you know what if you're physically near a store that they're trying to charge you more for that product because they know well you made the effort to actually go get all the way to that store so you must really want that item. Um or uh you know a whole there's we have a whole blog post on this sort of some examples of all of the different things where we're seeing sort of really targeted pricing happening based on you know are you riding a ride share to a certain community did you you know is your battery low and therefore they know like you really need to get this ride share or else you're going to be stuck where you are. I mean it's incredibly sophisticated but really sort of at the basis is you know we have it's not that we haven't known the threats of you know all of this data collection both in terms of how it can like there used to be the term we used web lining right um and now it's surveillance pricing but we've been talking about these issues for a long time but those things that that were threats are now actually technologically feasible the friction of actually doing this has gone way down. It's become much more feasible to actually do it. Um, and we also see that on the government side as well. All of this information collection from video cameras, from license plate readers is now available to sift through with AI. And so it's it's AI on top of and layered upon the the infrastructures that have been enabled to proliferate since, you know, the 2000s. Not because we didn't think there were threats, but because they're honestly a lot of uh, you know, we haven't been able to connect the dots as well for people. And so I think Corey, what you mentioned in sort of um earlier about people now realizing the real implications of this um you know earlier this year when license plate readers were really being weaponized in terms of immigrant communities and and deportations. I pulled out my talking points from 2007 that talked exactly about these issues. Like I did not want to be right. I did not want these hypotheticals to become reality. Um, but what is what is important right now is that people are really realizing what's really at stake. And this is a moment where the kinds of digital rights issues that many of us worked on 20, 30 years ago that were seen as niche, that people didn't really understand or realize the direct implications on their lives, their livelihoods, their personal safety, their economic security. um people now realize that and we have this incredible opportunity to actually build these coalitions to actually get these laws passed. Like there is no like so I go to presentations and I'm like you know most of the people we present to uh were not alive when the last federal privacy laws were passed in the 1980s or the 1990s. Um and you know you know uh our our privacy laws in the state level on the federal level uh should should not you know be older than most of the people who are working on those laws um as as staffers or um are are in communities voting. Um and >> well it's not like there have been any new privacy threats since Die Hard was in movie theaters. I don't know why we would need a new law. >> Exactly. And we we know we've needed new laws. I mean there was actually a real opportunity to I've been doing studying sort of of where where we are in history and sort of where we had these political opportunities and and what happened you know we had these political opportunities in the 1970s there was a there were really powerful forces that tamped that down in the late 1970s when we were at those early modes of computerization where we had the opportunity to really address issues of collection and use. Then we had another moment and we'd really made some progress. Um, you know, in the late 1990s there was, you know, opportunity there. Um, and then 9/11 happened and it was a very, I mean, we were around. It was a very hard political time to get anything through, let alone stop the things that were happening. um you know we and but we're at another moment now especially with AI like we are at a moment when we can where there's a lot of attention people are really realizing the kinds of things EFF and many others have been talking about about license plate readers and surveillance and surveillance capitalism for 20 plus years. Like that's why I'm so excited. Like the perils are also the opportunity. Like this is a moment where we can be like okay we have seen what happened before. we've seen the narratives that have undermined, you know, the coalitions and we can actually and that's why I was really excited for your book as well and all the work that you do because, you know, so much of this um can can [clears throat] be, you know, a little in the weeds for some people and you do such an amazing job of storytelling and we need those stories so people really understand what's at stake. So like when you write a book, you know, when John Oliver does, you know, his piece just last week on surveillance, like connecting the dots and bringing that to people, that is critical because you can go to court and you can go to court and if nobody knows about it, you may still win. But if you if people don't know the work we're doing in the legislature, in corporate advocacy, in community work, um we need people to know. We need people to care. We need people to push for that to push their legislators because we know what needs to happen, but we need the power to see it through. >> Yeah. So, that's very kind of you. And and you know, as we've been talking now, I've realized that we've been uh swirling around three of my favorite quotes from three of from one of my favorite writers, which is William Gibson, you know, maybe a patron saint of EFF at this point. Uh for those of you who don't know, William Gibson coined the term cyerspace and also once quipped, uh cyberpunk was a warning, not a suggestion, which I think describes so much of our work. That's the first quote. The second one is this quote that cyerspace is eververting, which is to say turning inside out. the the risks we used to worry about with surveillance pricing in your app are now surveillance pricing problems in the shop with electronic shelf tags with apps that you have to use to purchase things and so on. And then and then the the third quote that I want to pull out here and and you know move us on to the next section of the discussion is his famous quote from neurommancer the street finds its own use for things which is to say that when we seize the means of computation we can find ways to use the same tools that are being uh turned to oppress us to liberate us. And this is an area where I I think we've seen these advanced stat statistical inference tools that we call AI doing some very good work >> and I know we have uh an organization that we've done a lot of work with human rights data analysis group who've done some remarkable things with machine learning since the earliest days. They were the ones who very early on showed that um the predictive policing algorithms that were proliferating all across America were uh you know uh poised to sub substantially magnify the bias they were supposed to address that you know the the example they used using some very careful mathematics and some very good data sources is that if you have a police force that you suspect is doing racist stop and frisks and then you say okay well we don't want them to do racist stop and frisks we're going to ask the algorithm where they should go to do their stop and frisks and then you feed it the data from the racist stop and frisks, >> right? >> Then the algorithm is going to say, "Well, all the guns and drugs you found last year were in the pockets of the brown and black people you threw up against a wall in search. Therefore, that's where you're going to find them next year." And then, because that's the only place they search next year, the algorithm will become more and more convinced that all of the crime is is committed by uh racialized people. Not because that's true, but because garbage in, garbage out has been an iron law of computing since the 1950s. >> Exactly. >> And and this is a thing where we see various activist groups uh and an analytical groups doing very good work, using these tools, figuring out what's going on, uh finding ways to describe it, finding ways to demonstrate it, finding ways to test for certain kinds of conduct, all using these tools. or even just, you know, as we see with Innocence Project New Orleans, sifting through millions of arrest reports to find ones that are similar to arrest reports that led to successful exonerations, not because they're using the the AI to be their lawyer, but because the lawyer needs a sort algorithm to figure out which arrest reports to read first, and starting with the ones that look most like the ones that led to exonerations makes a hell of a lot of sense. So can you talk us through some of the ways in which you know the the we see less Empire Striking Back and more Return of the Jedi uh where where the Ewoks have figured out how to use the tools. Uh that's mixing a metaphor. >> Yeah. No, absolutely. I mean yeah I mean garbage in garbage out has been around like people all know this right and so in this iteration it's been sort of bias in bias out is what I've said you know Bibo um you know it is not surprising at all. Um I think one of the things I think you know AI is incredibly powerful and it can be harnessed for the things where we see a lot of the the holes. Um you know government never wants to evaluate whether or not these systems actually work because lo and behold when you actually do evaluate them they actually create way more harm than good. You know the police just like to say oh this surveillance technology is going to be a silver bullet. You know, I've been doing this since literally the pilot program of the first cameras went up in San Francisco because some mayor saw it in Chicago and believed it maybe would work. Um, >> magic beans, >> magic beans, snake oil, uh, every single ability when we've actually done research um, shows that, you know, lo and behold, it actually you doesn't address the problem and creates more harm. I've written like I think six reports on this and so many research but it's very hard to have those research projects take a long time and they are really intensive and so AI does create the potential for us to be able to sort that data to really sift through that to actually you know figure out um you know all that body cam footage that supposedly body cameras were going to be used as a tool for police accountability Haha. A lot of, you know, most of the time it's now been used for surveillance, but actually harnessing that back and being like, let's actually watch what the police are doing in that. Um, and and really flag all the ways that there actually has been, you know, use of force, um, improper actions. So, there's there's a lot of there's a lot of potential there. I will say I I went to an amazing workshop a year or so ago where there were a lot of public defenders there and it was really about having a conversation about how we can harness AI and utilize it actually for the people for protecting rights for enforcing laws that are supposed to be uh protecting the community from use of force and and ensure police accountability. I'll say of course that the resources are very lopsided for those uh that ability to do that. As you can imagine, the money coming from the government is mostly going to the police side or to the district attorneys. Much less of it is going to nonprofit organizations um or you know the public defenders or others that are actually pushing for that police accountability. But you know EFF you know has public interest technologists. That's an area where we are working in coalition and working with colleagues to think about how can we actually use um use technology not just AI but other technology to make sure that we're in the driver's seat and not just you know being pulled along for somebody else's ride. So, we're getting close to the time for questions and and you can stick your questions in the Q&A there. We have a roundabout process where those are then being put in another tab that I've got open on mine and we're going to try and get through them. Although, boy, there's a lot of them already. Um, but before we go to questions, I want to talk about what the people watching today can do. I almost said tonight. I I little spoiler alert, I'm in London. It's it's nighttime here. Uh, London, England, not London, Kentucky or London, Ontario. uh or London, Connecticut. Uh but um you know, we have a a a couple of you know, we have a couple of questions that our colleagues have asked us to address here. Um about what the people watching today and tonight can do. The first is what's one thing that regular people or local organizers should be building right now before the bubble pops so we're not caught flatfooted? Do you have thoughts on that? >> I mean, I have a lot of thoughts on that. Um, you [laughter] know, there's a lot of sort of action items on EFF's website. There's a lot of ways to get engaged, but I want to kind of go back to what you said. You know, wherever we live in the world, wherever we, you know, if we're in the United States, if we're around the world, if you're in any of those Londons, anywhere around the world, you know, these kinds of questions are happening locally. um in terms of how is AI being considered, how is it being potentially used, how is it being deployed, and I just want to go back to the fact that actually what we do locally really matters. Um and not just matters to our local communities, but can actually change the entire trajectory of what happens around the country and in the world. you know, the fact that um you know, in the United States, our real IDs do not have an RFID tag, that our US passports actually have to be scanned before they are read started because of local work that happened in a tiny town called Sutter, California. That changed the trajectory of everything that happened on that issue in the United States and on passports, not just in the United States, but around the world. And that's because somebody, you know, somebody felt like there was something that they could do and contacted EFF and the and at that time at the organization that I was at. That was work that I did with EFF in 2004 that is still protecting people. It's probably one of the biggest impact we've done that probably nobody knows what it impact really was. So one the local piece and then the other is to ensure that that local and state work can continue to happen. There's a huge preeemption fight right now. The federal government wants to pass laws at the federal level that one do not protect people and two wipe away laws that exist on the state and local level so that we actually can't address things in the future. So working on the local level and protecting what we can do in the local and state and in our in our in our uh countries around the world and also working together. These issues are not just happening in one community or in one country. And that's also I mean EFF works globally because these issues are interconnected. And so building our coalitions, doing our thinking, um doing our strategy together globally is incredibly important on these issues. Um and so I would say act locally, think globally. Um, >> so I I like the fact that this question talks about the bubble because it's not an area we've touched on, but it's an area that's pretty central to a lot of the work I'm doing because I think many of the of the most grievous AI harms will arise from the economic damage that's done from this extremely unsustainable practices of the largest firms that are, you know, burning a trillion dollars this year to make $50 billion in revenue. Uh, and you know, as Stein's law has it, a bedrock of finance, anything that can't go on forever eventually stops. And you know, if you're out there before the bubble burst trying to, for example, not have a data center built in your community without uh the regular processes of consultation and environmental review and so on. I got to say like your city government does not build a data center in secret because they think you'll be pleasantly surprised and they don't want to spoil it for you. They know you don't like this idea. That's why it's all happening without the regular processes. And one argument I rarely hear colleagues who are out there ma making these uh engaging these important campaigns about uh you know responsible democratic process and the planning of all kinds of u high-tech infrastructure including data centers is that this is a bubble >> and like it's not just bad that we're talking about like taking farmer Brown's cantaloupe field and seizing it through eminent domain and then building you know seven football fields worth of uh future laser tag arena. on it, it might never actually get switched on. You might just end up with a giant foundation slab where you used to have a farm if the bubble bursts before we get there. So, if you're looking for an argument to make before the bubble burst, if you're looking for a way to talk to, you know, city governments who are trying to do community development, who've been sold a story about the money that will come into your town if you allow the data center to get built, you need them to engage with the actual economics here of a sector with terrible unit economics that loses more money with every generation of its technology uh and that um is spending hundreds of billions of dollars to make tens of billions of dollars. a thing that just doesn't end well. Uh, as I think we've seen with lots of other bubbles. So, the last question before we go to the qu uh actually, you know what? I'm going to go straight to the Q&A because um we've got we're we're at quarter two um that maybe we'll we'll cover the last question here. So, um Nazarin uh wants to know uh what happens to democracy with a with AI in the hands of the government? uh they've asked a lot of questions but but um that's the one that I found uh uh quite trenching and related to our work. I have some thoughts but I'd love to hear from you Nikki. >> I want to hear your thoughts first. >> Okay. Well look I uh I read a stupendous uh article by a data scientist yesterday uh that I wrote up this morning for pluralistic.net that about the tendency of algorithms of all sorts that make predictions to have um predictions drive uh um outcomes that make the predictions come true. So Netflix predicts that you're going to want to watch one of 10 movies which it shows to you and you choose one of them. That doesn't mean Netflix chose the best movie. It means that you chose one of the ones that are recommended to you. And we see this with all forms of prediction and recommener system that this is um uh something well understood in all machine learning that when you preference exploitation over exploration when you make predictions based on the likelihood that it you've got a good fit for the criteria you end up missing out on all the solutions that might be better and then the subject of this prediction from this machine learning system enacts the prediction and makes it come true. And so one of the things that I'm quite worried about in the context of states is that states are asking AI to describe ground truth about our policy, right? Where what like what where are their problems, will these what solutions can we apply to these problems and so on. And so the government then takes the advice of the machine learning system. It says this person should be denied parole. This person's children should be taken away. this place should be zoned in a certain way, this legislation should be written in this way, and then that ends up turning into policy that vindicates the prediction. And just like the racist predictive policing algorithm that's, you know, sort of engaged in this kind of cooperia where it's eating its own outputs uh and gets maybe you could call it like pryion disease for AI where you're feeding the output of an AI back to an AI, you end up with increasingly deranged and unhinged policy. And and I think this is the consequence of this very opaque way in which our decision-making takes place. And I would say finally, although this wasn't in in this excellent paper, but it's something that Henry Ferrell, a political scientist, has observed is that there's something very seductive, especially for authoritarians about automating the civil service because the civil service is a source of friction. They're the people who know how things work. And you show up as a politician who's got ideas but doesn't understand the procedures. And you say, you know, I don't want to hear your excuses. Just make it happen. And the permanent civil service says you can't. And we see this with the Trump administration and, you know, the difference between one and two. And they just fire them all and replace them with chat bots, which was the the the project Elon Musk did for Trump. And um and then you have chatbots just pliably enforcing the will of people who don't want to hear about how those policies will play out on the ground. >> Yeah. I mean, you know, that's our Doge case, right? You know, all these issues. Um you know, and I wrote a piece last year that was about putting people power into privacy law and really, you know, defending and advancing rights and justice and democracy in the AI age. And so this question about sort of the intersections of AI and democracy, a well functioning democracy is a bundle of all of these rights and structures that are about you know making a bit we the people for the people by the people. Um and you know obviously we have been a very imperfect democracy right now at least in the United States and in many places around the world. there's been a lot that we have not done right even without AI and now we sort of are in this moment where you know how AI is used how it's developed how it's deployed is you know potentially going to even you know perpetuate exacerbate the types of inequality um the types of anti-democratic you know actions that are happening we see Trump fully utilizing it Um, and I think we kind of see it in in in in in various layers. I mean, I've already talked about government surveillance, it obviously supercharges, you know, the ability to know who people are and where they go and what they do. Um, so in in terms of the balance of power between the government and the people about our private lives and our ability to act and speak. So, we've got so the privacy bundle, we've got the free speech bundle, our ability to feel like and and safely uh engage, you know, communicate, connect, mobilize for justice. Um, you know, these, you know, I'm I'm a constitutional lawyer. That's where I' I've spent my career at the intersection of constitutional law and new technology. Um the implications of AI on our privacy and our ability to engage in free speech to mobilize for justice are profound. We've also got issues of I did a symposium this last spring where I really brought together sort of folks who traditionally more worked on sort of physical voting rights and then the issues of AI and how that intersects because you know many of us are familiar with Cambridge Analytica. I mean this is a whole much more sophisticated era of you know what how can you parse information? How can you then you know create um you know information in the voting context that's going to manipulate or you know move voters in a certain way. There's been some really good research on on those issues recently. So the intersections of AI and you know more sort of voting rights and manipulation um you know and uh impersonation you know obviously and that's that sort of comes back to the fact that EFF is holding all of these different pieces at the same time. There's some organizations that are just sort of working on voting rights or misinformation without sort of the privacy and free speech angle. So these are complex issues. These are nuanced issues where you have to understand sort of again not not putting your foot on one side that then pops up in this other way. Um you know that undermines our ability to actually uh speak freely because you've made a decision just about like trying to prohibit certain language related to voting that actually gets utilized for all this other precedent that is undermines our ability to actually speak. Um, so I think, you know, how does what happens to democracy? Um, uh, that's why I'm here. That's why I do this work is, you know, I care deeply about, uh, defending and advancing rights and the ability of people uh, to live our lives in safe and free ways in an emerging AI world. So, you know, I'm I'm here because I care deeply about us being able to fight for rights and justice and democracy in the AI age. Um, and there's a lot to do. Uh, and that's why I'm I'm so I'm so glad so many people have come to this uh to this uh live stream because it makes me feel like we've got a whole crew out there that we can actually be working with on these issues. And and I hope um everyone is also a member of EF because we need you. we need you in this fight. >> So, uh Mark Cunningham uh asks about uh the international situation um and uh and describes the EU as being far ahead of the US and who's capable of tackling this in the US. I I chose this question because it's an area that I'm quite interested in because I I was EFF's European director, our first international employee. Uh and I worked in in both uh intergovernmental forums like the United Nations, but also at the at the EU and Brussels. And we we now have people who are much better at that than me working in those roles. Uh, and one of the things that was very striking in the 31 countries I worked in when I was when I was doing this was that everywhere I went, I met people who said, um, we understand that the tech policy we have here isn't ideal, but the US trade representative has said that the quidd proquo for us giving America the tech policies that it would like us to have, including things like prohibitions on reverse engineering and very long terms of copyright and so on. uh and and even ironically one of the things the US trade representative has made a priority everywhere is saying you must not have the flexibilities in your copyright law that America has. You can't have fair use only we get fair use. You have to have much more restrictive copyright rules so that um our exports of culture, textbooks, what have you uh will not be flexibly used in your country. Well, all this was predicated on the idea that there wouldn't be any tariffs on the exports that those countries sent to America. And so Trump has inadvertently, you know, created a a different kind of liberation day than the one he intended. One in which there's just not the case anymore for other countries to do things. And this is very important because at the same time, Trump is wielding American foreign policy against um uh countries where they do try and reign in tech. So Canada and the UK proposed a 3% digital services tax to make up for the fact that all the tech companies pretend they're Irish except for Amazon, which pretends it's in Luxembourg. And so they don't ever and none of them ever pay any tax. So you can't get domestic competitors because they're giving 35% of their gross receipts to their tax uh authorities. And the US companies are not because of this weird Irish gambit. Um, and we've seen them crack down on rules about interoperability in the Digital Services Act and the Digital Markets Act. And we've seen American companies emboldened by this. Apple has refused to comply with the Digital Markets Act, >> uh, and has said, you know, go ahead and try and take it out of our hides. Uh, Trump won't let you. And so we're now at this very interesting moment where countries all over the world are starting to think about whether they can go into business exporting disinhitification tools, tools that reverse engineer and fix the defects in America's tech products. And I think this is an area in which Americans would substantially benefit because, you know, Americans, they're like the beta testers for every bad idea that some greedy jerk has in a Silicon Valley boardroom. And if you know Americans can figure out how to buy reasonably priced pharmaceuticals using the US Postal Service from Canadians, they can buy software to open up their iPhones or let them fix their tractors or their cars or put third party ink in their printers from someone in Europe or maybe Ghana or, you know, Mexico or any of these countries that have seen themselves in the crosshairs of the Trump trade policy. I don't know if you have thoughts about like what it's going to take to get American officials to to actually engage with these. I'm kind of feeling like um maybe we just let other officials solve our problems for a while if we can't get our own to do things. >> I mean, you know, I have always been a you know, a big believer in sort of thinking about overall systems and like where are the levers, right? Being nimble and you know EFF working at all these different strategies in the courts, in the legislature, in communities, globally means that we can actually see where those opportunities are at those moments and sort of create those plans that are both proactive but also take advantage of opportunities that pop up, you know, that might not be anticipated. And so, you know, this is not the first time that there are opportunities where another country is able to move something that then we can utilize in in the US. There's there's all sorts of examples of that in the privacy space where, you know, French regulators or Canadian regulators have been able to move issues that then we've been able to utilize in the US to put pressure on US companies, move that back. Um so to me it's really an entire systems approach and that's why you know I sort of I jokingly said you know work locally but think globally. It is about like how can we as EFF how can we in a sort of broader coalition really be working even more collaboratively with you know Canada the EU other countries um I'm excited for for that opportunity um because the issues that we are looking at are not just issues in one community or one state or one nation these are we obviously we you know the not only was the internet global But AI is certainly global and we're we're we need to be um that we are going up against formidable forces and we as an EFF community and as a coalition of organizations um you know really need to be working together to be thinking strategically and working together to address these issues and that's that's what I I wrote in my you know first blog post as executive director you know this is about working across strategy working across issue working across place to build the power uh for the future that we really uh want and deserve. >> So I want to close with a a last question here from biker Neil who says you know to we see these openweight models catching up with these closed source uh proprietary expensive systems. Do we see a future where AI runs on hardware that is under control of individuals, small groups and uh and that we we are able to access these tools without paying rent to corporate gatekeepers and and I would say I think both of us the answer is yes. But then he asked an interesting question which is much more an EFF question which is what policies and tech choices ensure that hardware makers, app stores, cloud platforms do not recentralize their control. what what can we do to support this uh a future in which AI like other computational tools are under the control of the people who use them? >> Yeah, I mean that's that's one of the things that's really on my mind and I'm excited for our public interest technologists to be working with others to really be thinking through those things. So I would say you know we obviously you know we built privacy badger we built surfbot we have been you know really integral and sort of the backbone of encryption on the web and this is an a new area that I'm really excited you know for us to be thinking about and so I'd say support support EFF make sure we have the resources to really think about those issues and dig into those issues uh because we really are bringing that expertise sort of across issue area and AI is, you know, we are living in this incredibly consequential time on AI and those decisions that we are making, the the systems that we are building, the climate that we are creating for AI to really be working for uh the people um is, you know, these are both, you know, the needs and real opportunity at this moment. So, I'm so proud to be the new executive director of EFF. I'm so proud to work with the amazing staff. Um, and thank you for being part of the community and uh looking forward to meeting you all in person and online even more in the years ahead. >> And and Sarah, before you close us out, I would just say that so many of EFF's core fights are the same whether we're talking about AI or or um any other kind of technology, the right to scrape the outputs of AI so that you can get your data out of them. I mean, these companies, it's amazing. They're like, "Well, when we said move fast and break things, we didn't mean move fast and break our things." Um, I I'm fine with breaking their thing. I'm not literally we shouldn't crash their servers, but like moving fast and taking your data off that stuff. That scraping is like that's how we move data from one system to another. Reverse engineering, the right to modify, the right to publish your modifications. All of those are as true whether we're talking about data center uh portability or app stores or any of these other technologies and they don't change just because we're doing a uh novel form of statistical inference with them. Uh foundationally if if you can't open it, you don't own it. Uh if if you can't modify it and fix it, then it's stuck. You're stuck with whatever ideas some person had before who didn't understand your needs. Nothing about us without us. we should all be able to have the finals determination about how our technology works. It's what EFF has fought for since uh the beginning 35 plus years ago. And it's a fight that we're never going to give up. So, thank you all for joining. And Sarah, I'll hand it over to you. >> Oh, Sarah's on mute. >> Oh, now it's now IT'S A LIVE CAST. >> NOW IT'S a live stream. >> Thank you. Thank you. That is all we have time for. I want to thank Corey and Nicole for joining us today. And thank you to everyone who contributed to the discussion in the chat and all the great questions and everyone watching around the world. We said it before and I'll say it again. Digital rights begin with you. EFF has been a leader in privacy, freedom of expression, and innovation for 35 years. And we need your help to keep that fight up. Take out 10 minutes now to become a member today at eff.org back/supportchange. And please continue these conversations about your civil liberties with your colleagues, friends, neighbors, and family members. And once you're done renewing your EFF membership, be sure to mark your calendars for our upcoming segment of Affecting Change: Elections 2026: Polling Places and the Risk of Surveillance. Thanks for joining us and see you next time.