← The Lawfare Podcast
Loading summary
Transcript107 lines
- [00:01]
Renee DiResta
Need real insight from industrial data Versit with a single source of everything and get the best outcomes.
- [00:11]
Grow Therapy Announcer
Transform the everyday with Siemens History shows women keep showing up for everyone every day. But who's showing up for you? Grow Therapy helps you put your mental health first with therapy that's covered by insurance and built to support you whether it's your first time in therapy or your 50th. Grow makes it easier to find a therapist who fits you, not the other way around. You can search by what matters like insurance, specialty, identity or availability and get started in as little as two days. There are no subscriptions, no long term commitments. You just pay per session. Grow helps you find therapy on your time. Whatever challenges you're facing. Growing GrowTherapy is here to help. Grow accepts over 100 insurance plans, including Medicaid in some states. Sessions average about $21 with insurance and some pay as little as $0 depending on their plan. Visit growththerapy.com acast to get started. That's growthherapy.com acast Growth acast availability and coverage vary by state and insurance plan.
- [01:27]
Scott R. Anderson
You know folks, when I started the Lawfare podcast, it was kind of terrifying. You know, I didn't really know how to podcast anything. I had no idea if anybody was going to listen to it. What if people think it's dumb? What if I make an idiot of myself? Over time the thing has proven the test of time. It has a listenership. But it was a really scary moment and I know it is for others and that it helps when you have a partner like Shopify on your side to help.
- [02:04]
What is Shopify?
- [02:06]
Shopify is the commerce platform behind millions of businesses around the world and 10% of all E commerce in the United States. From household names like Allbirds and Mattel and Heinz to brands that are just getting started. You can tackle all the important tasks in one place, from inventory to payments to analytics and more. There's no need to save to multiple websites or to try to figure out what platform is hosting the tool that you need. Everything's in one place, making your life easier and your business operations smoother. So it's time to turn those what ifs into with Shopify. Today, sign up for your $1 per month trial at shopify.com lawfair go to shopify.com lawfair that's shopify.com lawfair.
- [03:13]
Tyler oh, what is that?
- [03:16]
Molly Roberts
Sanzo Calamansi Lime. The best flavor.
- [03:20]
Scott R. Anderson
What is this drink?
- [03:23]
Molly Roberts
It's the bougiest seltzer you could possibly conceive. Of it comes from Whole Foods. It surely comes from elsewhere, too. It's authentic Asian flavors made with real fruit.
- [03:37]
Scott R. Anderson
Nothing suggests that it's not real fruit more than authentic Asian flavors, I'm not gonna lie. So I'm glad they clarified that.
- [03:43]
Molly Roberts
Yeah. Not just as flavors. Authentic ones.
- [03:46]
Tyler McBrien
I hate to admit it tastes really good.
- [03:48]
Molly Roberts
It's so good. It's so good. But it's. It's. It's. Somehow they made Spindrift even more expensive and even more obnoxious.
- [03:54]
Scott R. Anderson
Give us the product placement. Let's go for it. Maybe we'll get some ad revenue out of this. Throw it. Throw it out there. What's the name? What's the name of the brand again?
- [04:00]
Molly Roberts
Sanzo. I would love to be sponsored by Sanzo.
- [04:02]
Scott R. Anderson
That Sanzo hit us up, Sanzo.
- [04:05]
Molly Roberts
Save me a lot of money to be given Sanzo for free, let me tell you.
- [04:14]
Scott R. Anderson
Hello, everyone, and welcome back to Rational Security, the show where we invite you to join members of the lawfare team as we try to make sense of the week's big national security news stories. We are getting in the weeds on this week's episodes. We have a couple of stories that have been percolating, a few of which have just broken literally in the last hour or so that we have been tracking, we think will be interesting to talk about, even though they may be a little bit a step beneath the headlines, at least until this news broke this week, in the last hour or two that people may not have been following as closely, but we think they're pretty interesting things that we should be paying attention to. But I'm thrilled to have a set of my colleagues here to talk about them with me. First off, joining us for the first time in a while, thrilled to have her back, lawfare contributing editor Renee Diresta. Thank you for coming back on the podcast, Renee.
- [04:55]
Renee DiResta
Thank you for having me.
- [04:56]
Scott R. Anderson
And it is, in fact, your segment, the one I'm going to lean most heavily on you for, that has now broken news. So we have very timely things to discuss. Thank goodness. So we'll all be scrambling and no doubt making many errors as we discuss exactly what just happened in the last hour. So don't hold us too accountable, listeners. It literally just happened. We'll get to that in just a minute. Also joining us live, managing editor Tyler McBrien. Tyler, thank you for coming back on the podcast.
- [05:18]
Tyler McBrien
Thank you for having me.
- [05:19]
Scott R. Anderson
And now an old standby on the podcast, officially having made more appearances in her, what, six months or so with lawfare than it's fair to come to a little more than that. Six, seven, eight months now is LawFair senior editor Molly Roberts. Molly, thank you for coming back on the podcast.
- [05:35]
Molly Roberts
Thank you. Thank you for calling me old, old, reliable, stalwart.
- [05:39]
Scott R. Anderson
None of it sounds great when you say it, but they're all good things in this context is how I mean them. There's just no nice way to say it. But we apprec you coming back here on the podcast as we break into a couple of big national security news stories, starting with Topic one the Metaverse of Madness. On Tuesday, a New Mexico jury reached a $375 million verdict against Meta after a seven week trial that focused on whether the social media company knowingly harmed children's mental health and facilitated child sexual exploitation through its algorithms. And just before recording, another verdict came down in a jury trial in California about whether Facebook and YouTube have proven too addictive in a way that harms an individual plaintiff in case several other similar civil cases are set to go to trial in the coming months. What do we make of these verdicts? Do they signal a turning tide against social media companies for the algorithms that make them both profitable and addictive? Or at least potentially addictive? Topic 2 Saving Face President Trump and Republican congressional leaders went back and forth this week over a deal that would put forward a bill to fund the Department of Homeland Security, or at least big, less controversial parts of it, Despite President Trump's threats not to sign any pieces of legislation until Congress passes his Save America Act. Trump views the Save America act as vindication for his criticism of the 2020 election. Republicans in the Senate have hedged and resisted his calls so far to blow up the filibuster in order to pass it. Instead, they now appear to have a deal in place that will allow less controversial parts of the funding for DHS to go forward and for the funding for the most controversial parts, particularly ICE and removal operations, to go forward through reconciliation on what is likely to be a party line vote along with select chunks of that Save America Act. All this still yet to be finalized. This is the plan as we understand it at the current moment. Why is President Trump so determin to pass the Save America act? And what does the compromise he now appears to have reached with Senate Republicans mean for its Future? And Topic 3 Poly Wants a Crack Up Flight monitors, Pizza Place Trackers and Google Earth the past few years have brought open source intelligence, better known as Osint, into vogue. Accounts on X have racked up millions of followers by quote unquote, monitoring the situation for news events spanning from Russia's invasion of Ukraine to natural disasters. But this explosion of OSINT accounts has brought a wave of disinformation and coincides with the growth of online prediction markets such as polymarket and Kalshi, whose betters use OSINT to gain an advantage and at times to manipulate the results. How has OSINT contributed to the online media landscape and how has it heard it? So for our first topic, Renee, I want to come to you. So we literally did, just as you can tell from me, stumbling over my effort to adapt the pre written paragraph to describe and set up the topic. We just got news on this just a little bit that we've been struggling to catch up with. We know two days ago this judgment came down, it's New Mexico jury trial. Literally just in the last hour or so, another verdict has come down in this Los Angeles case, both in state court, both advancing somewhat different but related theories of liability under state law. Talk to us about these verdicts and why they represent a potentially significant new avenue or opening for accountability or liability, whether you think it's properly accountable or not for social media companies that we haven't seen at least as clearly capitalized on by plaintiffs in the past.
- [08:50]
Renee DiResta
Right. So I think one of the things that we're seeing here is a shift from lawsuits that focused on the content to the dynamics, the product design specifically. Right. The conduct, the argument being that the product design and then what the platform chooses to recommend, how the platform chooses to engage people is moving more into the realm of platform conduct as opposed to just adjudicating content disputes. So the case in New Mexico was for listeners who are not familiar with it, the state's sort of smoking gun was something that they called Operation Metaphile, where undercover agents set up accounts posing as 12 and 13 year old girls. And then within a month, one of these, you know, fake children had about 7,000 followers, almost all of whom were adult men. And then the kind of kicker with this particular situation was that instead of meta systems flagging the suspicious adult activity, it instead started to send the child automated tips on how to monetize their account, grow their following, you know, do the sorts of things, put them on the influencer path. So the jury found that that was a kind of a consumer protection issue. Right. That platforms were telling parents that these platforms were safe while internal data. And they also had a couple of internal whistleblowers who testified internal commentary showing that inappropriate actions, I think something like over half a million inappropriate actions with minors were happening every Single day. So that was the scope of the New Mexico case. So this is. This is not a question of, you know, is this a user's speech that needs to be protected? This is the platform nudging people in particular directions. When I worked at Stanford Internet Observatory, we actually did a research project where we got a tip about underage users, real underage users in this case, who were marketing their content on Instagram. This is illegal. And so one of the challenges there, you know, we did reach out to the company, Obviously, we do responsible disclosure. But one of the things that was challenging about that was once you engage with a few of those accounts, it pushes you more. The users and the others. The dynamics of the recommendation engine are such that the platform is actually actively nudging and making suggestions to users to do certain things. So I think this question now that we're getting into and in the Los Angeles case, was about mental health. Right. Content that is pushing teenage users in particular to. To see things that are intentionally addictive is the dynamic. Right. So it's. It's an intentionally addictive product, and then the content they are pushing may also be harmful. This is two different facets of the question. So I think, you know, what we're seeing here, it made me think a little bit about. They're talking very much now, these court cases, about defective product features. So almost like a car having bad brakes, right? You know, when an algorithm chooses to push a specific video to a child to keep them addicted, the platform isn't just a passive conduit anymore. It is actually the designer of the harmful experience. So that's where we're starting to see things. The comparison here, I think you're seeing come up next, actually a metaphor that I think Tristan Harris and others were using almost a decade ago at this point, is that they're more like tobacco companies than just public squares. And that's the shift that we're seeing in the decisions, the findings, and also the legal approaches to pursuing in that direction.
- [12:03]
Scott R. Anderson
It is a really, really fascinating development. I mean, we have followed Section 230 questions, which is the big set of immunities that has always been the bare barrier for accountability for these companies for a long time. And that's, you know, focuses on providing immunity for the actual underlying content, which has always been interpreted fairly broadly. So here the question becomes, when can you draw the line between the harm as what stems from the content versus the way the content is accessed, packaged, and delivered essentially to the user? Strikes me as the distinction.
- [12:34]
Renee DiResta
Yes.
- [12:34]
Scott R. Anderson
Which is. Can be a Fine line in some of these cases. So I'm kind of curious about like breaking it down across these three sorts of avenues. One we have the one that's the Los Angeles case where it's clearly the case where the plaintiff says, you know, I got addicted to this. I don't even think in that case it's necessarily clearly harmful underlying content. But it was because of the mode of delivery, the type of delivery that it was, I think particularly the. What is it, the endless scroll feature that I think was specifically a feature in Meta that at least got picked up in the media coverage. I don't know if I imagine because of YouTube was involved as well. Other things were in that, in that case as well, like that sort of design feature is what fed to the addiction, obsession, kind of compulsive behavior. Then you have in the New Mexico's kind of two related theories. One is that you actually had a case where they were either the product design sort of failing like you said about you were actually like, or a way you were actually contributing the structure of design contributed to these negative outcomes. And another one was actually specifically about the marketing under state law in particular, where essentially they were trying to point out that Meta had said they met certain safety standards and were safe in certain words, and they didn't quite get there. So for the Medicaid in particular, I guess my question for you is like both these cases are unique because they at least particularly the New Mexico case I'm less sure about the Los Angeles case seems to have heavily leaned on both this Meta files investigation you described and the fact that you have a lot of whistleblowers who come out about Meta specifically. We know for a couple years now that Meta has a very messy and complicated record with safety and its relationship to safety, despite public statements to the contrary. And that we've had whistleblowers bring information to Congress to other venues about this that we saw featured in play here. Is that something that's replicable with other social media companies? Does that make Meta uniquely vulnerable to these claims? And like, do we have a sense about how important that is to the viability of these claims?
- [14:32]
Renee DiResta
I don't think that Meta is going to be all that unique here. I think a lot of them have very similar types of features. I think per your point, maybe the question is what do they say about them? But is the marketing copy like cleaner for TikTok? Maybe, I don't know. But that that dynamic that these are safe places for teenagers, These are perfectly fine. There's nothing Addictive, the research is wrong, et cetera, et cetera. They've spent a lot of time. This is where I think the tobacco comparison also comes in. Right. Merchants of doubt. The idea that, oh, you know, you can't prove that this is bad, you can't prove that this is harmful. Now you're starting to see, I think, juries and others finding that in fact they are. There was an interesting case on the CDA230 thing. I was actually trying to see what the latest on it was, but there was. It was like the Anderson vs. TikTok case, which said TikTok's recommender system might actually constitute the platform's own speech and that it could be held liable for content that it promotes as it represents first party speech rather than just hosting third party content. So where's the 230 line there? But that is another. That was another one that came to mind as we were, you know, thinking through what are the shifting dynamics where you're starting to see kind of chips in the wall, so to speak, that has held back some of the accountability for a very, very long time.
- [15:52]
Tyler McBrien
Scott, I didn't realize you were a plaintiff.
- [15:55]
Scott R. Anderson
It's a big Klan, for better or for worse that people hop in there. Yeah, I mean, that really gets like. The fundamental question here is like, when do these algorithms become something independent of the underlying content? The metaphor, I can't remember. I came up with this, or it was in the oral arguments during the Gonzalez and Tomna cases a few years ago. I remember I thought of this in the context of those because I almost used it in a piece and didn't. But it's kind of the idea of like the serial killer notes you see in movies where people cut individual letters out of magazines and piece them together in a note. In theory, is that just third party content being packaged a particular way, or is that actually individual speech? And that line, while clearly there are opposite ends of that, and we can see the two pillars of it where the trade off is exactly. Can be hard to draw.
- [16:38]
Molly Roberts
Well, that will invariably come up with AI content too.
- [16:41]
Renee DiResta
Yeah, exactly 100%. That's why I'm very curious to see how some of those cases come out, particularly where, you know, the AI is actively defaming people and things like this, or giving actually harmful product, you know, giving harmful medical advice, giving harmful health advice, harmful food advice. Actually, I got some pretty remarkable tips on what happens when. What do you do with rotten food? Well, you can wash it. You know,
- [17:05]
Molly Roberts
this is like the rocks. Pizza stuff too.
- [17:07]
Renee DiResta
Yes, exactly. It's moldy. That's okay, just scrape it off, you know. But no, I think so that, that piece of it is. It's kind of unclear, I think what is going to happen, but I was actually pulling from Meta's widely viewed content report for a talk that I'm about to give. And one of the things that they put out that's actually very interesting is over time, the dynamics around widely viewed content. You know what, what people are engaging with, how they're recommending. There's a little, they're always put in a little pie chart. And then the little pie chart, it tells you where it's sourcing the content from. So as it is creating your feed, what percentage roughly people are seeing from their friends, their family, news organizations, things they follow? It's at 41% unconnected at this point. And I remember last time I delved
- [17:54]
Scott R. Anderson
into unconnected means, like outside of.
- [17:56]
Renee DiResta
Oh, sorry, unconnected. So unconnected. No, no, no. Unconnected content means you don't follow the account, you have no relationship to the creator, you're not, you haven't opted in, essentially, you haven't formed, you know, you're not following me and then seeing my content instead. 41% now is the platform intuiting what you want to see. Right. It's almost like the TikTok ification of all feeds, because that was TikTok's secret sauce. The idea that your social graph actually didn't matter all that much. Recency things other people were engaging with popularity, a whole bunch of different ways to weight what it's selected could actually be a much heavier driver of engagement than just saying, these are the people in this person's social graph. Let me show them posts from their friends. So I think last time I pulled up the widely viewed content report and really delved into that particular stat was when we were writing this paper actually on AI slop and why the recommender system was because AI slop was starting to be in the widely viewed content report. And I think around that time it was two years ago, it was about 26% was unconnected content. So again, a pretty significant shift, if I'm remembering correctly, of that, that evolution, 41% actually surprised me how high it is now.
- [19:03]
Scott R. Anderson
Totally, totally.
- [19:04]
Molly Roberts
But I guess the platforms could say something even when you talk about unconnected content. Like, well, we're still not that different from a bookstore that decides to have books to put on the shelves. Right?
- [19:15]
Renee DiResta
Editorial.
- [19:16]
Molly Roberts
Yeah, right. And so that's what I wonder with all of this. And I'M kind of heartened to see that the protections that Section 230 provides are not now being interpreted to be as absolute as they were a few years ago. But also I'm surprised and a little bit skeptical and I wonder where this is going to go because again, I just see the companies saying, well, we're still doing content moderation. Our algorithmic choices, our design choices are just contemporary content moderation. And that's exactly what section 230 lets us do. It says we can't be punished for doing it. The whole point of Section 230 is that it's a sword and a shield. It lets us do moderation without making us be in trouble because we moderated. So that's what I'd imagine they'd say. And I think that's a tough argument when it comes to something like you have endless scroll and people are suffering from the mere fact of being addicted from being online, if that can be proven. I think it's harder. And this is what you were kind of speaking to earlier, both of you, Scott and Renee. I think it's harder if it's. Well, you were online and you saw content that promoted body dysmorphia, because then that is speaking to the content. The content is just downstream of those algorithmic choices.
- [20:25]
Scott R. Anderson
Yeah, it is really tricky to figure that. I totally agree. I think that's the spectrum of arguments we're going to see. And we should note, like New Mexico has another bundle of arguments they're going to make, which is actually like really directly inspired by the tobacco cases, which is the idea of Meta being a public nuisance or creating a public nuisance that's going to allow them to pursue, I think, substantially more damages. This was a $375 million judgment. So like, I think that's the number big scale judgment. But that's because they had a huge number of cases, each of which was capped at only $5,000 of damages. So it's kind of like a volume thing. I think they can go for much more of a public nuisance and importantly they can do which they've done with tobacco companies, pharmaceutical companies, things like that, and really push for remediation plans and Meta to fund programs that aren't just providing damages and particularly like fairly nominal statutory damages, but in fact actually like have to address the underlying problem, or at least that's what they they successfully done in those other prior cases where that fits in. I don't know, because I don't know exactly what the legal standard of a nuisance is and how much of it it depends on. I think their theory about how much of it ties back to the underlying substance of the content versus your choice to steer it. But I do wonder whether like this gets into what was in the Gonzales and Tomno case. Like there the issue was well, look, we have fairly what the I think the justice described as fairly neutral or agnostic algorithms that yeah, they do favor and push things to certain but that's based on our assessment of what they're looking for, their search term, their past history. It's neutrally agnostic towards any one particular content or interest stream. But inevitably it will weight things different ways depending on how people engage with it. And there the court ultimately made its decision based off some wonky stuff about the Anti Terrorism act, but did seem to suggest, and I believe even in the final opinion did suggest this isn't the sort of stuff that at least in the ATA context, maybe any and notably the ATA context is like basically imports common law tort standards. So it's a good reasonable guidepost for that. That's not the sort of thing that you understand to be reach the kind of mens rea requirement of aiding and abetting, of knowing and intentionally doing something. But like I think the question becomes if you have evidence and you know evidence of like systematic trends of steering like children towards sexual predators, right. Like does that cross a line at some point where it is knowing and involved in there is this harm and you become aware of it. They said in that case like they tried to point out evidence that there was knowledge on the part of companies that this did sometimes happen. The companies then come back and yeah, it did sometimes happen, but we actually actively took measures to counteract it. Maybe not 100% effective all the time, but it was still result of this broader kind of neutral program, not a deliberate steering. But at a certain point that does seem to break down. You can only claim neutrality so far if you know it's steering things towards these particularly harmful outcomes. I think. But I don't know, it is an easy line. I think those claims are harder certainly than the things that are so intrinsically about the design features like the addiction, which really could be addiction to any, any sort of completely benign underlying content, but it's being fed to you in a super problematic sort of way. Renee, where do you see the companies responding to this or acting like this? They're obviously going to challenge it. It's obviously going to go to the court. They're going to go all the way Supreme Court. If people get their druthers about it. And the Supreme Court had feelings about this. So maybe the companies don't ultimately want to bring it to the Supreme Court because they think they're worried they may lose. But if they don't, that's going to go to at least the state, probably Supreme Court eventually to resolve some of this stuff. And because it's got the section 230 hook, they very well may bring it to federal court. So how do you think they're responding business wise? Set aside the appeal question. Like they're going to have to live with this at least for the next year or two while these appeals are ongoing, maybe longer. And they have a lot of other cases coming down the pike. Does this affect their business model? Does it affect how they do things?
- [24:15]
Renee DiResta
I mean, it's an interesting question. They had actually scaled down trust and safety operations rather significantly. It's going to be interesting to see if they scale some of that back up. Meta made some announcements about using AI to moderate more. It's going to be interesting to see how they're going to again, is the public going to trust that? I think the public sentiment also is a very big deal here. People increasingly do not trust them. This narrative that they're harmful for kids is very, very prevalent. And I think that is one of the key challenges that they're also going to face. Just that public opinion dynamic is going to be heavily influenced by these verdicts as well. So what they're going to do to make themselves seem like good faith players when they've actually really swung very, very hard in the opposite direction away from content moderation as they allowed that bullshit political campaign that the hard right ran to reframe it all as censorship. They're going to have to walk that back in some way. And now it's going to be interesting to see how they do it.
- [25:08]
Scott R. Anderson
Yeah, it does seem like they've put themselves in a difficult, disadvantageous position, especially if you're, if you're a defense to your. Like our algorithms have all these problems is that. But we have other algorithms embedded in AI that will try and fix it and enhance it. Like it kind of layers back and I think that kind of leads like the last issue that's worth talking about here. What does this tell us all about AI liability? Potentially in my mind this seems like it opens the door to a lot of more this model of liability around AI, where AI at its core is kind of repackaging content you've heard people raise. Well, maybe you can make section 230 defenses about this. Certainly when it's acting like kind of like more like a search engine or research bot, you may have stronger arguments there. It becomes, becomes harder when there's not attribution, when it's really actually repackaging, reselling. And that is ultimately what AI is doing a lot of the time. Is this the sort of model we're going to see, this kind of like product liability? I mean it seems like it is, it seems like those are the sort of cases we've already started seeing in the AI context. But I wonder where there's an intersection here. Maybe even if the prevalence of AI, the fact that how much people talk about and think about it in the last year or two compared to, you know, when Gonzalez and Tomno were decided, when all these cases were last really heavily debated at the highest levels, maybe has people a little more sensitive to that kind of algorithmic harm sort of argument.
- [26:27]
Tyler McBrien
I'd also add, I mean these, these really visceral, horrible stories of AI models directing, seemingly directing people to self harm or to kill themselves. I mean it seems like they be like, like it's a lot easier to attribute the harms to an AI model that is like directing you and exactly telling you how and why you should rather than just being kind of fed algorithmically. And like it's been grabbing headlines I think for good reason because some of these chat transcripts are just so disturbing.
- [27:03]
Scott R. Anderson
Yeah. And like where does that liability adhere? Is it when the chatbot like refers you to a manual to like, like commit self harm, in which case you're protected by section 230 if they pursue the acts in that manual? Or is it in the kind of coaching leading up to it? And here we seem to see a line dividing there, which is that it says it's really the latter, not the former. But you know, where that line is in practice could be tricky. In a variety of cases, I suspect,
- [27:28]
Molly Roberts
and I do think partly because of AI, we've now entered the age where people are kind of afraid of their phones and afraid of the world that is coming, where younger people who maybe aren't afraid of their phones are going to be in kind of this virtual reality world all the time talking to AI instead of talking to humans. And that's extremely troubling and scary and maybe has made people who are already skeptical of the social media companies get even more serious about the perceived harms.
- [28:00]
Scott R. Anderson
Well, speaking of efforts to save Americans, let us go to the Save America act that is now being debated in Congress. President Trump Trump has had a bee in his bonnet about the Save America Act. This is like predominantly, as far as I can tell, I will not pretend like I've sat down and read through the whole thing. An elections reform bill aimed at doing a variety of things, many of which are responsive to not very well substantiated claims about 2020, many of which appear to get about certain bugaboos that people have had about elections and things like voter ID for a long time in certain circles, combined with a couple of new measures we've seen the president face down primarily against his own Senate Republican colleagues around this, because the filibuster is the big barrier to this particular bill, or at least a lot of it, although notably very thin margins in the House from terms of Republican control as well, which could also pose a problem, although it can pose a problem both directions, really. We've now got this compromise solution that looks like parts of it may at least be submitted to a majority vote in the reconciliation process, which bypasses the filibuster, although only if they can get past the Senate parliamentarian and the notorious birdbath process through which all these reconciliation bills have to go through. So, Molly, tell us a little about what the president's broader ambitions are with the Save America act and where they appear to have been winnowed down to, and what happens, I guess, both to the rest of the bill if this reconciliation package is the vehicle by which Senate Republicans say this is all we can do on this.
- [29:29]
Molly Roberts
Yeah. So Republicans say election reform, Democrats say voter suppression. That's generally what's going on with the conversation about the Save America Act. I think that, you know, President Trump has said that the outcome of the midterms hinges on this act. Whether that's really true, whether the particular reforms that they want to push through would very clearly be to Republicans benefit everywhere, I don't know. But that seems to be the perception, and it's not that difficult to believe because essentially what this is mostly is a bill that makes it harder to register to vote and then harder to vote by requiring you to show your papers in various ways. So there's the showing your papers of proving you're a citizen, and there's the showing your papers of providing photo id, but super specific types of photo ID that are not possessed by most Americans. And so the people who'd be most likely to be effectively disenfranchised by that would be low income people, minorities, and also particularly harmed would probably be women who have gotten married and changed their names, because then you have a mismatch in what your birth certificate says and what your ID says. So, yeah, looks like that would benefit Republicans, probably. That's what Trump seems to be saying, or what Trump is saying. But also I think kind of separate from what the act was actually substantively do is the idea that this is a narrative that says elections need to be saved. Right now, elections can get rigged. And so even if this doesn't pass, you know, Democrats have rigged the elections if they win, because we needed this act to make Republicans have a fair shot. So I think those are the reasons that he cares so much about it. As far as what's happening now, it's a little complicated. It ties into the fight to fund the Department of Homeland Security. There's a lot more urgency now among the Republicans to do that because of the situation at the airports. That is making the shutdown of the department way more visible to Americans. Now people are actually getting upset at the Republicans. So they want to fund dhs, but the Democrats don't want to fund ice. So there's a compromise budget deal kind of in the works that would fund all parts of DHS except ICE's removal operations. And the question for that is, will enough Democrats go for it? Because they're not really getting all the reforms to ICE that they want. And there's this sort of weird situation where it's like, okay, if they're willing to fund more of ice, maybe they can get more reforms. If they won't fund any of ice, maybe they get no reforms. So that's where that conversation is. But Trump said, I don't want there to be any compromise unless we also get the Save America act done. And that's what brings us to the reconciliation conversation, because what it's looking like is going to happen is that Congress is going to try to pass this compromise bill and then take care of the parts of ICE that haven't been funded later in reconciliation. Okay, that's not all that surprising. Has to do with spending, but they're also going to try to get portions of the Save America act through reconciliation. The problem there, of course, is that the Byrd rule says that anything you're doing through reconciliation has to be related to the budget. All the stuff I just described about the Save America act doesn't have anything to do with the budget. So how do you tie it into the budget? That's where we get to kind of the winnowing down that you're talking about. So I'll stop for now there.
- [33:06]
Tyler McBrien
Yeah, maybe I can throw another Question your way or anyone's way. I'm curious whether you see this, the Save America act more as a last gasp of this so called Election Integrity Movement. I think a big question that people had in the last presidential election is when Trump won, where all of this energy and this organizing and this movement, like where it would go, would it just dissipate because Trump won or is it, did it actually. It had so much momentum that it had to be. There had to be some sort of outlet. And is this what we're seeing or is this actually like the movement is picking up strength and this is just the first big proposed bill of what's to come? I don't know. I guess like whether the Electoral Integrity movement, even when Trump wins, it doesn't go away way.
- [33:58]
Renee DiResta
Trump is not the Election Integrity Movement. Just to be clear, Trump is the fake, the election is stolen bullshit movement. Let's just be really totally candid about that. And everything that is happening is him trying to optimize his chances. It is transparently political. We all know that. I just want to point out that
- [34:11]
Scott R. Anderson
he says it, he says this is.
- [34:14]
Renee DiResta
Nobody is hiding the ball here. That is not a, that is not a partisan statement. That is just a recantation of like, of what is actually quite plainly said. In fact, if you go and you look at, at Cato or Heritage, particularly Heritage, you can, you can look at their database of non citizen voting incidents. And I think it was 24 instances that Heritage found between 2003 and 2023. That's what we're dealing with here. We're not even in the three digits on, on Heritage's database there. Right. So the idea that this is an Election Integrity Movement is bullshit. And I think that we can't use that frame. I think we need to be constantly reinforcing. Whatever you think about voter, about voter id, that is a completely separate question. Then is there an epidemic of illegal voting in our elections? The answer to that is categorically not.
- [35:03]
Molly Roberts
Yeah, well, I agree with that totally. And I think that to me the Save America act is kind of frightening for two reasons that go back to the second point of him doing it, which is to create all this doubt and create the perception that we have this election integrity crisis, which as Renee said, we don't have. You know, I think if the act passes, it's dangerous not only because it might disenfranchise a lot of people, but also because, okay, yeah, this will probably benefit Republicans. Even if it were a wash. The stricter the rules are, the more ways you can get someone disqualified, the easier it is to say when you lost that you lost because a bunch of people who should have been disqualified were able to vote. So that's something to me that's super scary about it. As to whether it is the last gasp of the election integrity, not integrity of the, of the so called election integrity movement, the self proclaimed election integrity movement, I don't think so. I think they're gonna try to do this in a lot of ways. I think that if they can't do it through Congress and we can go back to whether they'll succeed through reconciliation, but I think it's gonna be tough to do anything meaningful there. States are already doing this on their own, some of them at least. Also, it's possible that the President and this is the subject of a piece that Anna Bauer and I wrote for Lawfare and we did a podcast about it as well. But it's possible that the White House would try to achieve at least some of this on its own, though that would be almost certainly unconstitutional. Essentially, I think there are various ways they're going to try to get at least pieces of this done, some more realistic than others.
- [36:42]
Tyler McBrien
ActiveCampaign is the marketing automation platform built for big swings and big dreams with intelligent suggestions, powered by AI and your data, generate ideas in seconds, import your brand and create full campaigns with simple prompts set personalized messages backed by real time feedback, smart segmentation and effortless reporting that tracks every win. Let's redefine what's possible together. Get started for free@activecampaign.com Over 90 of
- [37:13]
Scott R. Anderson
the top 100 US accounting firms trust Bill to handle bill pay processes.
- [37:18]
Tyler McBrien
Why?
- [37:18]
Scott R. Anderson
Because our tools are built on over a trillion dollars of secure payments. We're not just moving money, we're powering financial workflows for half a million customers. That's a level of expertise you just can't fake. Ready to talk with the next one? Visit bill.com proven to get started and grab a $250 gift card as a thank you. Terms and conditions apply. See Offer page for details.
- [37:42]
Grow Therapy Announcer
History shows women keep showing up for everyone every day. But who's showing up for you? Grow Therapy helps you put your mental health first with therapy that's covered by insurance and built to support you whether it's your first time in therapy or your 50th. Grow makes it easier to find a therapist who fits, not the other way around. You can search by what matters like insurance, specialty, identity or availability and get started in as little as two days. There are no subscriptions, no long term commitments. You just pay per session. GROW helps you find therapy on your time. Whatever challenges you're facing. Grow Therapy is here to help. Grow accepts over 100 insurance plans, including Medicaid in some states. Sessions average about $21 with insurance and some pay as little as $0 depending on their plan. Visit growththerapy.com acast to get started. That's growththerapy.com acast growtherapy.com acast availability and coverage vary by state and insurance plan.
- [38:54]
Scott R. Anderson
And I'll say like this is a lot of these proposals have origins that well predict Trump or 2020 election. I mean these are things that I remember talking about when I used to do voter protection work, like in 2008 and before that even to some extent like their proposals that have always been out there dominated by myths with some kernels of reality about whether it's a handful of cases, whether it's like there are certain administrative inefficiencies that in theory might lead to situations but that in practice there hasn't really been any evidence of it being any sort of large scale or meaningful or pivotal sort of change. And what the trade off is, I mean that's the real question here. Like how much do you value absolute security of elections if it deters people who can vote but don't have to vote from being able to do so? And it's a values trade off. It's one of the things where there actually isn't a right answer. I know what I think the right answer is, which is that I think more people should vote who should be eligible to vote. And unless there's a real problem of fraud, we shouldn't be wasting time deterring them. But it is ultimately like one of these value trade offs. And I think we're just going to see continuing cycles of that. The one thing that's really interesting about this though, I think is that 2024 was the election that when it came out of it showed that a lot of voters that are vulnerable to some of these measures, older voters, Hispanic voters, voters with like lower education rates who are on various surveys show they have less likely, are more likely to have limited access to ID information that they need to be able to vote. Donald Trump carried a lot more of that vote than prior Republican candidates have. And that was a conversation coming out of 2024 that I remember very vividly because I'd been involved, was volunteered a little bit during the election and came out. Remember coming out of the election, people saying this really looks like such a different sort of electorate and pattern in this case. And I do wonder whether that is if anything, a sign of maybe, maybe not the death knell of these claims, but a sign that they're maybe getting a different sort of reception or emphasis because the way that they cut is. Isn't that clear? I think a lot of times the way people assume the political impact would be was actually based upon a lot of broad assumptions about who votes for each party and that the actual reality is, and there's. I've been kind of scrolling through and looking through studies from bred and center and a few other groups show that like really strongly suggest that the impact is going to be really all over the place depending on like local elections, what population is a particular area is, how it affects it and what the nationwide impact of that be. It can be hard to judge. So I mean, maybe that's one of the things that has people, maybe Republicans in the Senate a little more nervous about this because you don't really know what the outcome is. And maybe it's better to deal with the flawed system you have. But, you know, and that got you elected if you're an incumbent than one that you don't.
- [41:38]
Molly Roberts
Yeah. Or at least skeptical enough to not want to blow up the filibuster over it. Right.
- [41:42]
Scott R. Anderson
That's a better way to describe it. Yes, exactly. Exactly. Which is really interesting. So where do we think that this leads to Donald Trump in his broader 2020 motivated campaign? So, Molly, you've talked about some of this stuff might come through executive order. We can talk a little bit about that. That's one avenue. I don't think there's a lot of other legislative chances if this doesn't get through. Because if they get this DHS bill through, there doesn't seem like there's anything else that they're going to have if the president could not persuade them to break the filibuster over this sort of stuff, it doesn't seem like they're going to get the filibuster broken over other stuff. Seems like enough Senate Republicans are still on board with that. So if that's where we are, where does he channel his energies and concerns about this elsewhere? Or does he finally have to accept, which something that reportedly a lot of Republican advisors and supporters of the president have been telling him, which is that this is a political loser and we need to drop it.
- [42:36]
Molly Roberts
I don't think he drops it.
- [42:38]
Scott R. Anderson
I tend to agree that seems unlikely at this point. He's got a lot of skin in the game on this one?
- [42:44]
Molly Roberts
No, I mean, I think he. If the midterms don't go his way, blames it on the failure to have the Save America act passed and says that we tried to save the elections and they wouldn't let us save the elections and they stole the elections. I think. I think that's what happens.
- [42:59]
Scott R. Anderson
Yeah. Yeah. It's interesting. I mean, like, it is one of these catch 20twos of this whole process that. Because there's like, you know, you're trading up these values, you can make those arguments no matter what the actual underlying rules are. Because you could say, well, these people were able to vote who shouldn't have been able to vote. In theory, all these other people, like, you can say these arguments because it's all kind of a relative system. But, like, you know, other than broad, in the absence of widespread evidence of actual fraud, it just seems like the availability of those arguments doesn't necessarily mean they should be persuasive. It's just something that people are going to be able to make. And if people are motivated enough to believe in it, that's kind of the decider. I just don't know if we can plan around the ability to make these arguments that much one way or the other.
- [43:40]
Tyler McBrien
Before we move on, I feel like I should go on the record and say I don't think this is about election integrity.
- [43:47]
Scott R. Anderson
Good clarification, Tyler. So there's one more part of this, which is, of course, whether all of this is even really feasible. And that comes down to this question of the birdbath and the reconciliation process and the politics surrounding that. So, Molly, tell us what your sense of this is. Is this a compromise that's actually going to go anywhere, I guess, either on the DHS front or on the Save America front?
- [44:09]
Molly Roberts
Yeah, I was pretty amused to see that Senator John Kennedy of Louisiana earlier this month said that what they have to do is they have to hire a really smart lawyer to figure out how to make this surv. Birdbath. But I think that was also in a speech or statement where he said that the Karen wing of the Democratic Party is in ascendancy. So there were a lot of kind of funny phrasings in it, but I think you do need a really, really, really, really smart, perhaps genius lawyer to do it, or at least do it in any meaningful way. So basically, what you have to do and what it seems like Republicans would try to do, is incentivize, pay states to change their rules, to do some of the things in the SAVE Act. But that's a far cry from the binding national that are really supposed to save our elections. It doesn't seem nearly as impressive, if that's all that you manage to achieve, is to incentivize, is to have states do this voluntarily rather than forcing them to. And it seems like what's circulating, what they're trying to incentivize states to do is implement voter ID laws, require proof of citizenship for voter registration, and share some voter data with federal agencies for verification and then post election audits, which makes sense because part of the idea is we can say the election was stolen after the election. So it seems like that's what they're trying to do. But again, the only way that they could even possibly do it is to suggest to states, hey, we'll give you some money, you do it on your own. And even then, it's not clear that that would survive the Senate parliamentarian scrutiny. At which point the question becomes, well, do the Republicans want to just ignore the Senate parliamentarian? But they haven't shown much interest in that. And, and that's basically, if you break the Byrd rule to that degree, it's basically like blowing up the filibuster, which they don't want to do also. So I think that's where it leaves us. I don't think that many Republicans would even vote for the reconciliation bill that included these provisions of the SAVE act if they survived the parliamentarian scrutiny. Susan Collins doesn't seem super into it. Lisa Murkowski doesn't seem into it. It. It's a tough thing to ask them to vote on, particularly because these reconciliation bills tend to have a lot of other stuff added onto them that can often be quite controversial. So here it might include social safety net cuts, it might include a bunch of money for the Pentagon, and so just controversial provisions that many Republicans might not want to vote on in this party line bill. And that would make it even harder for the SAVE act stuff to get through if it makes it that far.
- [46:43]
Scott R. Anderson
Yeah. And one other tool they seem to be trying to work in Save America provision through is through the conditioning of different grants and additional grant money. So providing money and then saying, if you want this money, we're going to have to like, make it conditional upon things like documentary proof of citizenship for getting voter registration. And that's one place where it might have some teeth. But I know those provisions. This is where I wish our colleague Molly Reynolds could be on the podcast, unfortunately, out the other Molly R. Because she of course has literally written the book on this. And that is the sort of thing that I know has run into birdbath problems in the past. I can't 100% remember exactly where the line is. So I know in some cases we've seen certain things like that get through, in other cases not. I don't know how well defined it is or how we know whether this gets through. And that's why you need not just a lawyer. Actually, you don't even really need a lawyer. You need a parliamentarian who understands Senate practice about this, which is, I will say, as somebody went to law school, not something you learn in law school to any particular extent unless you really go down some odd paths. So yeah, we will have to exactly where this leads. But certainly it is an interesting legislative development to watch in the weeks to come. So speaking of watching and predicting things to come, let us transition to our third topic, which is all about the future. Predicting the future, understanding what's going on around the world, and then potentially sometimes even making money on that future. Tyler, you wrote a really phenomenal piece in the Baffler that I highly recommend to folks. And I don't have the title in front of me, which I meant to have, and I accidentally closed that window, but you can tell us in a second why, when I hand the mic over to you, that's worth reading. All about the kind of flowering and mass proliferation of OSINT as a practice on social media. In some ways that have been good. I know you've taken interest in osint. I have relied on OSINT practices and producers at various points in various moments. But that may have reached a critical tipping point in the other direction, at least in certain contexts. Talk to us a little bit about your article and the trends you're observing in that and some of the positive, negative, negative implications we're seeing coming out of it.
- [48:44]
Tyler McBrien
Yeah. So I called it situational unawareness.
- [48:48]
Scott R. Anderson
There we go. I can't believe I forgot that. I'm embarrassed now. Okay.
- [48:51]
Tyler McBrien
It grew out of. I'm sure many listeners have also seen these dashboards that have just exploded on mostly on Twitter that are essentially promising this immense surveillance capacity that you can just have on your browser. They're styled after a Bloomberg terminal. Or they say, you know, you can have like Palantir in your pocket or you can basically promising the capabilities of an intelligence agency. But in reality these are just, I think vibe coded AI slop that just Frankenstein together. Ticker symbols, maps, different like newswires, different Twitter accounts. And it just becomes this, this cacophony that's pretty useless unless you have contextual knowledge, if you have, if you know how to just read the signals and the noise. And it just seemed like this really silly thing at first, but then kind of a dangerous proliferation because a lot of them look good and therefore seem authoritative and they sort of give this illusion that you know what's happening. And then I think it became really dangerous when this marriage started to emerge between these OSINT style dashboards and then prediction markets. A lot of prediction markets now have their own dashboards. So there's one thinking of called Polyglobe, where it grew out of the Pentagon Pizza index, if some listeners are familiar with that, which is this idea that an increase in traffic among pizza restaurants around the Pentagon is a good proxy for there being a buzz of activity in the Pentagon. And they're up all night planning military strikes so they got to eat all the pizza at night. It's, I think anyone with knowledge who follows would understand.
- [50:41]
Scott R. Anderson
Sometimes you feel like Chinese and that's what you order in on your strike nights.
- [50:45]
Tyler McBrien
You need a Chinese restaurant monitor as well to supplement. But it has just, I mean it was kind of like a little joke on Twitter for a while and these tweets would go viral that there was an increase in activity before a certain event. But obviously the tweets that showed an increase in activity when no event happened happens don't go viral. But then it just has been taken to this crazy level and then like I said, the marriage with prediction markets started to really worry me which, which gave rise to this, this piece. But yeah, so I, I, I can, I can leave it there for now. I'm happy to, to talk about any parts of it. But yeah, I encourage listeners to go look at some of these things because there's so many of them out there now. And you must be a crazy person to think that you can kind of glean any insights, any actionable intelligence from these, this just mess of data.
- [51:36]
Molly Roberts
I was in a small town in Maine and I walked into this shop and the woman who was middle aged woman at the counter was being really chatty and asked where we were from and we said DC and then she started talking about the pizza index. So this just shows exactly how far this thing has spread and she was really sold on it.
- [51:55]
Tyler McBrien
And some of it is kind of funny. I mean the one that I was talking about, about there's one connected to the pizza index and they have like a tiered threat level and instead of defcon they call it Dokon and it's like very kitschy. But then it just gets very unsavory when it's. Then you can't. They have a disclaimer at the bottom about the use of this data. But it was literally, when I was scrolling down, it was literally blocked by a polymarket banner ad encouraging you to go there to place bets in the prediction market. And the idea obviously is that you can have an edge in these markets because of all these streams of data that you're monitoring every second. And it just seemed like this mix of doom scrolling and AI slop that just felt very of the moment. And then I'm sure people saw that Polymarket had this pop up in D.C. at a bar where you could monitor their situation in real life. And this was a manifestation of, of a very common type of tweet that I've seen over the past few months, which is like bros, imagine a bar, but all the screens are like Bloomberg and et cetera and maps and whatever. So they actually did it. And I think much to a lot of people's delight, it was a complete fiasco. And a lot of the monitors were not working on the first day and it was this sort of like Fyre Festival esque thing that blew up in the face of a often maligned company. At least right now.
- [53:25]
Scott R. Anderson
I have to say I saw the pictures of the event which some of our colleagues went to, I know, to do a little on the ground reporting. I have never seen a less fun looking party. It's not entirely fair because of course they're using photographers. I think knowing the story is going to have a slightly critical event, are using full flash photography when it was probably a dark bar room night. It's kind of like at the night at 4am in the club where they turn all the lights on. Everybody's like, ah, this is shocking and terrifying.
- [53:48]
Tyler McBrien
I want this to deny.
- [53:49]
Scott R. Anderson
Yeah, exactly. But it is very entertaining because I look at these pictures, I'm like, wow, this looks terrible even by my standards. And I am a 42 year old father of two and I do not get out. I would not come to this party, which is not a good sign. So Renee, you have spent obviously a huge chunk of your career thinking about disinformation, how it intersects with the Internet and prediction market. And prediction market, yes, yes. And related issues and where they intersect. Talk to us a little bit about how you make sense of this. There is obviously some degree of value in the lowering of barriers. Entry to.
- [54:21]
Renee DiResta
Oh yeah, yeah, yeah. For sure.
- [54:23]
Scott R. Anderson
Out there. Like Lawfare is a benefit of that in a lot of ways, right, like lawfare gets to do stuff now that you had to be a newspaper to do 15, 20 years ago, right. Or a law firm. We're able to do a lot of it cheaper and frankly, it's getting cheaper and cheaper by the minute between AI and frankly, a lot of OSINT sources that if you do them, you know, properly vet them, but then you run to the risk of this information slop. How do you manage this as a consumer and maybe even as a policymaker? Like, how do you approach differentiating these things?
- [54:48]
Renee DiResta
So so much of it is incentives, right? I was in a. Back in the day, I'm trying to remember the year now. It would have been maybe 2017. 2017. So almost a decade ago now. Gosh, we had a slack called Data for Democracy and a lot of it was OSINT research, right? Because it was sort of the early days then 10 years ago, where it was the recognition that you could get really interesting signal from stuff that people were just putting on Twitter, right? There would be like, remember when ISIS was a thing, even like five years before that, you'd have these like, jihadis who would like, not turn off the, like geotagging on their phone pictures, post a picture and like, boom, you've just got a ton of information from that. You could do triangulation of where people were. Bellingcat emerges, right? And this is, I think, like the pinnacle of fantastic osint. And you can start to see how these independent investigators actually can pull together quite a lot of things. I mean, some of the work that we did on Russia and the Internet Research Agency was exactly that, right? It was people thinking like, hey, I think I've got a signal here. Anybody, anybody here want to look at this with me? And then there's this very collaborative process, very fun process. Candidly, I love investigations where you're in there just trying to figure out what is the reality of the situation on the ground, what can you learn, what can you find? Even prediction markets too, Even. Even before that. So going like five to eight, 10 years maybe before that, I was part of this thing called the Good Judgment Project, which was, I think. I think it was maybe funded by DARPA or it was one of these kind of original OG efforts to see if super forecasters and others could actually do a decent job at predicting world events. And there too, I think. And also there was this question about experts versus crowds. There's a whole lot of different things that went into the idea of prediction markets being markets. When I was at Jane street, we would talk a lot about that. Right. How can you get as much information out there, people putting their information together? I think overall, again, it can be really fantastic. But the question is incentives. Because what you start to see is the rise of the fake OSINT accounts on X, right? The explicitly manipulative ones that are actively trying to put garbage out there to, particularly around the time when Russia invades Ukraine, right. To mislead people about what's going on there. You start to see them using even AI generated content at times to kind of flood the zone. And that question of how do you differentiate, I think is a broader information environment question at this point. It's not just the OSINT accounts that are, you know, participants in this ecosystem, but they're seen as being kind of authoritative, right? They're seen as being synthesis accounts. And that is where I think you start to get in some interesting questions around, particularly because they're also often anonymous. Right? Bellingcat is interesting in that it very much is not right. It really shows its work. So I think to answer your question of how do you know if it's good? The question is, do they show their work? Are you racing to get something into the media? Are you racing to profit from it? Are you racing to blast it out? Or are you going to do a very methodical investment investigation, put it out and like walk people through essentially a replicable set of steps. The ability to kind of gamble on it is where you do start to get to unethical things like markets being placed, where then people go and actively try to shift the outcome. And that, I think is the other thing that we're talking about here.
- [57:58]
Tyler McBrien
Yeah, I totally agree with Renee that a lot of it comes down to just kind of fundamental media literacy things. But specifically at osint, the show your work thing, I think, you know, anyone trying to figure out if it's a legitimate OS recent effort or not is they often will publish a very lengthy methodology and you can see step by step and so it could be replicable. And also this just more of humility and just some recognition of the limitations of what they're doing and corroborating multiple data points instead of just saying, drawing one very authoritative sounding conclusion from a single video. No one who's actually trained in OSINT would actually do that. And then the other thing is that a lot of the more legitimate accounts essentially have their own beat. You know, there's the guy who tracks shipping and he knows shipping, knows which ships there are and he is very knowledgeable and he doesn't really go beyond that Lane, because a lot of OSINT is. It looks easier. It looks a lot easier than it actually is. It takes a lot of technical knowledge and a lot of like, just, you know, subject matter expertise that is often not displayed in, you know, the end product product.
- [59:05]
Renee DiResta
And when we did investigations at SIO, like we, you know, early on in 2019, we found some pages. My colleague Shelby Grossman, who teaches a great OINT class herself, we found some pages that we thought were Wagner Group, right? So Y Prigozhin, operating in Africa, did a whole big internal investigation. We did reach out to the platforms. Again, responsible disclosure. Hey, you guys, is what we're seeing. What do you think this is? Right, so the attribution is done jointly and then before publication, I absolutely reached out to Bell and Cat and I was like, hey, I just want, I want another pair of ey on all of this, right? And you reach out to a bunch of different, as many different people as you can possibly get, honestly, who have the time or bandwidth to look over your work and make sure that you're not going to put out something that's just going to be wrong or embarrassing or if you're going to mention somebody, Right. If you're going to attribute an account to an operation, you're making a real statement there, right? You're saying, this is not what it seems to be. This is not authentic. Sometimes there's a real person attached to that, they're using their real name and you're saying this is linked to this other thing. And I think in order to do
- [60:02]
Scott R. Anderson
that, not to mention libel risks and other risks.
- [60:04]
Renee DiResta
Right, exactly. I mean, there's so many risks that go. You know, you would be so stressed before the night before the report comes out. You're just like, you read every sentence 5,000 times. I can't even tell you. So watching people just yolo it is kind of remarkable to me.
- [60:16]
Tyler McBrien
The other thing that's been really, really bugging me is these tweets that I'm sure a lot of people have seen from polymarket, or it's mostly polymarket, sometimes Kalshi, I think that they're tweeting like they're a newswire where they'll say breaking and then some very. Just one sentence, very simplified version of something that probably actually did happen. But it's. There's no attribution, there's no source, there's no further context. It's just a single statement. And then they followed up with a reply tweet that's like, here are the grand odds for this thing that's related. And it's just. It seems like very, very, very toxic. And then you look at the numbers that these tweets do and it's like in the thousands and sometimes over a million views and doesn't seem good.
- [61:00]
Scott R. Anderson
Yeah.
- [61:00]
I mean, it strikes me as kind of like the. Like a version, a mass kind of wave version of kind of the inshification process. I think it's Cory Doctoro's like, term. Right. Whereas the whole idea is like you had at least some of these platforms, like take the pizza one that I think you use as a really compelling example. Tyler did note an interesting correlation and documented it and was like maybe a data point. Like maybe it's not 100% right. Nobody's saying anything is 100% trend, but made it a point. Then it got popped popular to the point that people even in rural Maine are talking about it. Then it started getting a lot of clicks and then it wanted to make money off those clicks. And then all of a sudden you start getting banner ads and these poly market input and it becomes, by the way, potentially less effective. And it's trumpeting its own effectiveness much more prominently because it needs more clicks, it needs more money. It's this whole incentive structure, like you talked about, Renee, about what actually delivering it is. But the challenge, I do feel like there's fundamentally a real difficulty that I'm curious about what the era of AI and other things dragging this will actually may make this easier or harder about verifying even the methodology, like the credibility that you all tie to, which is that a lot of people aren't in a position to actually assess the methodology effectively. Right. You actually need level expertise to even do that. I'm reminded of. I don't know if you guys ever been to the Museum of Jurassic Technology. It's like my favorite place in the world in Los Angeles. It's like a museum dedicated to oddities in theory. But it's really, I think, like a big examination of the construction of knowledge and authority and how they relate. Cause it presents everything as like a museum. So it feels very authoritative. And it's presenting these completely oddity things, like a bat that can fly through a wall. And in that particular exhibit about this bat that can nominally fly through a wall, it gives you this, like, very old school, long, dry, boring. I remember it being like a deep German accent. I'm not sure that's 100% right. Like, description of the methodologies these naturalists pursued in studying this bat and the details of this bat and then like eight minutes in or something. I think I'm exaggerating a good way into this recording more than any normal person would ever listen. It says, and then the bat flew through the wall and you're like, oh, okay, okay. I have no way of knowing if that's true. It has all the outward indicators of authority, right. It's in a museum. It has this fancy scientific voice describing this. What sounds like a reasonable methodology to a layman, but it's a real challenge. What does AI mean for this environment? Does it mean. On the one hand it means, yeah, you can gather and synthesize all these things. Does it also mean we can stress test it a lot more effectively or at scale? Is that actually, actually a spot where AI may be a real virtue in a way that doing things at a scale and effectiveness that like lay people just can't.
- [63:32]
Renee DiResta
Meaning, do you mean using AI tools to do the investigations? Like for one thing that's interesting, there are certain aspects like geolocation and other areas where, you know, I will occasionally ask an AI to tell me like what's happening in this picture? Where is this? Where would you, where would you guess this is? Just again, not because I'm going to use it in a professional full follow through thing, but I'm just interested in where the capabilities are at a given moment.
- [63:55]
Scott R. Anderson
Tracking where your children are. I understand, I'm getting there myself.
- [63:58]
Renee DiResta
Yeah, we all need to know, you know. No, I think it's actually, it's also, you also want to know like which ones are good, right? Who's, who's got the best, you know, best, best technology for doing these sorts of things. Then there are actually the, the rise of AI detection companies that are trying to, that offer services that will tell you if something is AI or not. But those are an interesting mix also I think, because you have both false positive problems and false negative problems and the resulting challenges. Confidence. I think teaching people that there's sort of like a confidence interval here, that these are not actually binary determinations. A lot of the time it's giving you a guess, right? This is how likely we think this is. We did a project looking at fake faces on LinkedIn and tracking a network of inauthentic accounts on LinkedIn. They were AI generated, generated accounts. And we partnered with a detection company and we're like, this is, here's our list of profiles we believe to be inauthentic for the following reasons. Can you run through these images, right, and give us back the. Basically the assessments, and they come back as scores. That's how we're given that information. So when you have these detectors that just say to the public, AI or not AI, I think it's actually obscuring. It's trying to make it simple. But in making it simple, it's actually obscuring some of the information that is actually most useful to people who are doing serious analysis, which is how confident are we actually in this? And that is a real issue. Because the other problem, though, is that people do want to know. And you're starting to see this phenomenon of, you know, the. Grok, is this true? Right? And they're at Grok, is this real. At Grok is this AI, right? Which is basically trying to use grok to do. Again, the thing that I was talking about, which of these models are adept at detection? And the answer is, like, honestly, not very many of them. So they're using the tools that they have available to them, but they're not the tools that are right for the job.
- [65:53]
Tyler McBrien
And you know what else is binary? The outcomes of betting on the prediction markets. And so that also throws it in there too. Of you need to know, yes or no. Did this event happen so that there can be a payout or not?
- [66:04]
Scott R. Anderson
It's a really fascinating space and one that intersects with our work in all sorts of ways. I'm sure we're going to have opportunities to revisit it. We are out of time for this week's episode, but this would not be Rational security if we did not leave you with some object lessons to ponder over in the weeks to come. Tyler, what did you bring for us this week?
- [66:21]
Tyler McBrien
I brought a shameless plug. I've been working on a podcast series with our friends at Goat Rodeo and the Atlanta Journal Constitution that is finally out. The third episode came out on Tuesday, March 24th. It's a series, so there's a few more episodes in store. It's called who Blew up the Guidestones? Yeah, you can check it out any podcast platform or ajc.comguidestones I can say
- [66:47]
Scott R. Anderson
from personal experience that I saw Tyler take physical ailments in the form of poison ivy infestations, I think, and bed bugs and assorted other ailments that we mostly documented here on Rap Social Security in the travails of recording this podcast. So it is well earned and there's a lot of blood, sweat and tears going into it on Tyler's behalf, if not others, among other friends of the Lawfare family at Go Rodeo. So really Excited. I've just checked out the first episode. It's great. It's a great listen. I'm looking forward to the rest of the series. So definitely check that out. Molly, what did you bring for us this week?
- [67:18]
Molly Roberts
So, yeah, the prediction market stuff and the betting. I mean, my understanding is that 90% of the volume is sports betting, but obviously a lot of what we've seen that I have found to be distressing and in extremely poor taste is voting on stuff like, when will we bomb Iran? And so I brought this, this old, like, vintage don't ask why I have it Strategic Air Command hat that says, nuke them till they glow. It says more than that, too. It says, and use their asses for Runway lights.
- [67:51]
Scott R. Anderson
I increasingly think you live in a very esoteric TGI Fridays because you have a wide variety of, like, weird household objects that you're bringing. Object lessons. And I do appreciate. This is why I keep having you on, Molly, because I'm eventually going to run out of cool household objects you have lying around. Unless you're speed ordering these on Amazon or something.
- [68:08]
Molly Roberts
No, I am. Am. I'm simply going down to the war room and taking them and bringing them upstairs. Like, that's genuinely what I'm doing. This is that. But anyway, I think it captures both the kind of vibe of, oh, let's bet on the end of the world, and let's bet on people's lives and deaths, and also perhaps the sort of administration attitude toward, let's show how big and strong we are by starting a war in Iran.
- [68:38]
Scott R. Anderson
Yeah, fair enough. Fair enough. Well, since we've been talking about gambling, I thought I would throw in. I'm going to throw in a more normal recommendation. I'll throw in a local one just because I know the local ones aren't that useful for other folks. The more normal recommendation. Although I think this piece has gotten a lot of play, so I feel a little ridiculous making my jacket mystery. This is, of course, I think, the piece a lot of people have been talking about, the McKay Coppins piece about sports gambling, which seems very apropos of this particular conversation. Sucker is the title. That's actually not. I always saw it. I got it through my little social media as my year as a degenerate gambler. But I guess Sucker is a formal title. And I pulled up, this is in the Atlantic. It talks about how he spent a year doing sports gambling on the Atlantic's dime and ultimately came out concluding that he was essentially addicted to sports gambling. I was sent this article by every living person. I know because I occasionally dabble in sports gambling. For the record, very small stakes, not that often. And actually I didn't even do it once this past season, but the last few seasons I did a little bit. But you know, it is not a great story about this phenomenon of gambling that is now not just about sports but about all sorts of things in our life lives. I will repeat the joke I told my wife when she emailed this to me, which was that what he didn't understand is that year two is where your luck really kicks in. But not true, not actually true. It's a really interesting piece worth checking out about the broader phenomenon. As for my local recommendation, I realize I don't think I've actually said this and it's one of my favorite places in D.C. so I'm going to throw it out there just to throw a local business that I went to five times for some reason the last week and a half that I really, really value. And that is the best bakery I've ever been to, which is here in Washington D.C. called Saylou s E Y L O u over on 9th Street. It's phenomen. Don't know why I used to live right next to it didn't go nearly often enough. Now I live halfway down the city and I still go there all the time. I went five times this past week. It's phenomenal. It's great for every possible occasion. Breads, pastries, whatever you got. They got pizza on Wednesday nights. That's phenomenal. Definitely check it out if you're in the D.C. area. So I'll throw a little love towards Sailu Sailou. I'm a beloved customer and I have a special place in my heart. I feel a little bad I haven't thrown you a recommendation here on the podcast among all my other local haunts here in D.C. in the past. And with that, Renee, I'll turn over to you. What do you have for us this week?
- [70:37]
Renee DiResta
Mine are so late. I'm in a hotel room in Portland right now. I literally. Like, I could show you a pillow.
- [70:41]
Scott R. Anderson
I don't think in Portland's that late. It's a very hip city.
- [70:44]
Molly Roberts
It is.
- [70:44]
Renee DiResta
It's very hip. I'm gonna go outside and find myself some donuts after this. It's lunchtime here and I was in. I was in Madison for the last couple of days. I had a brandy old fashioned for the first time. That was very interesting. That was my. Yes, I've never had one of those before. But no, I feel like I've been, I've been working on work stuff. The only the thing that's been dominating in our house that my kids have been texting me about nonstop is like the New York Times pushed me. An entire entire like six article series on this was like the the BTS reunion in Seoul. The the giant K pop concert that happened. And so I am on the road getting deluged with K pop concert clips from my children. The stray kids dominate movie is in theaters and live streaming. The BTS one is coming. So yeah, between work being on the road and, and then this little glimpse into family life, that's really all I got.
- [71:36]
Scott R. Anderson
Hey, we'll take it. We'll take it. What lies ahead for any of us with small children perhaps is a big K pop future. And that is fine. That is fine. Even Hunters is getting me faced into it. Well, regardless, that brings us to the end of this week's episode. Rational Security is of course a production of Lawfare, so be sure to visit lawfaremedia.org for our show page, for links to past episodes, for our written work and the written work of other Lawfare contributors, and for informational Lawfare's other phenomenal podcast series. While you're at it, be sure to follow Lawfare on social media wherever you socialize your media. Be sure to leave a rating review wherever you might be listening. And be sure to sign up up to become a material supporter of Lawfare on Patreon for an ad free version of this podcast among other special benefits. For more information, visit lawfaremedia.org support our audio engineering producer this week was me of me and our music as always was performed by Sophia Yan and we were once again edited by the wonderful Jen Patcha. On behalf of my guests Molly Tyler and Renee, I am Scott R. Anderson. We will talk to you next week. Until then, goodbye.
- [72:41]
Renee DiResta
Want to turn your timeline into a fast lane digital twin it to outpace the field with fast and confident decisions. Transform the everyday with Siemens.