Ctrl-Alt-Speech

Close Your Apps and Think of England

• Ben Whitelaw & Jen Weedon • Season 1 • Episode 107

Share episode

Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.

0:00 | 44:30

Become a Ctrl-Alt-Speech supporter to get extended episodes of the podcast plus the chance to submit stories for us to cover.

In this week’s roundup of the latest news in online speech, content moderation and internet regulation, Ben is joined by Jen Weedon, a T&S veteran of Meta and Niantic. She is currently consulting and teaching at Columbia school of International and Public Affairs. Together, Ben and Jen discuss:

And in the extended episode for Patreon supporters, they cover:

Our fun links this week are the How Alberta eradicated rats (Ben) and Mogwooooo’s Instagram account (Jen).

If you’re already a Patreon supporter, you can get the extended episode on Patreon.

Ctrl-Alt-Speech is the podcast where we make sense of the major debates shaping online speech, platform power, content moderation and the future of the internet. It’s co-hosted by Mike Masnick (Techdirt) and Ben Whitelaw (Everything in Moderation).

Ben Whitelaw

So Jen, you've worked in the industry long enough to know now, and been on the internet presumably long enough to know that it's not always fun. But we try on Ctrl Alt Speech to, ahead of talking about its, you know, failures and frailties, to bring a bit of light relief. What fun things have you seen on the internet this week? what has brought a smile to your face?

Jen Weedon

So I can share, something that my son and I have created for the internet that has brought smiles to our face. So we have an absolutely preposterous looking cat. It is a Siberian, which has a lot of hair, and he has a very kind of, sour expression all the time, and very blue eyes, and just a lot of shenanigans. And so we've wanted to share sort of the adventures of, his name is Mittens, but we call him Mugru. Uh, we've wanted to share some of his, adventures, and so we've created an Instagram account, which I felt conflicted because he's-- my son's eight. But it's opened up a lot of opportunities to talk about sort of online dynamics and, and whatnot, which I can share more about later. so we have an Instagram account for Mittens, who also, it also, has cameos from our other, ancient cat, Oatmeal. and recently they've been enjoying our raised garden bed, so we've been gardening. So I have cat content, we also have a chicken coop that they go in, so it's all sorts of whimsical cat content.

Ben Whitelaw

Oh my God, so delightful. This is, this is a better answer than I could even have hoped.

Jen Weedon

Yeah. Well, I will share, I will share his handle, although I'm very keen to not make him obsessed with engagement metrics. So more just let's share the joy of Mittens in a creative way. my most recent post this morning was, likening Mittens to Michael McDonald because they both have-- I don't know if you know Michael McDonald from the Doobie Brothers.

Ben Whitelaw

Not f- not that familiar.

Jen Weedon

I will share with you. both of them have very blue eyes, and both of them love, the Doobie Brothers, so.

Ben Whitelaw

Love it. I think the internet needs more cat content, and so, you are providing a public service there. I- interestingly, my, fun internet story this week is actually about the cat's sworn enemy, I would say, but, a mouse. I, I've been blighted by a mouse this week that has been kind of somehow, found its way into our, our London house. I'm sure there are many, many mice running around London. and so I was very interested to read a piece by, Works in Progress, which is a kind of quite cool magazine that's been, actually been bought by Stripe recently, about rats in Alberta, Canada. Did you know, Jen, that there are no rats in Alberta, Canada? Not a single one, apparently.

Jen Weedon

where have they gone? Have they been eradicated?

Ben Whitelaw

They have... There's been a kind of systematic public program to get rid of rats because of a kind of public health, edict back in the 1950s, and I think a concern about the impact on the local agriculture. So in Alberta, will not find a rat. if you see a rat, you report it to government authorities, and people will, will come out and, and address the rat. You cannot keep a rat as a pet, which is, you know, maybe bad for f- if you're, like my neighbor from when I was younger who had a rat. so they're very, very strict on it, and, it also means that they, the Albertans can't really, don't really know a rat when they see one. There was a very funny stat that said that there was 875 reported sightings of rats in Alberta last year. Only 47 were real. So, so clearly the program is so good you just don't know what a rat looks like.

Jen Weedon

Interesting

Ben Whitelaw

yeah. I thought so. I thought so. So there was... the beauty of, internet publishing in particular brought me that story at a time when I needed it most, I think. A bit of solidarity with, my mice catching.

Jen Weedon

Have you been able to eradicate your own rodent?

Ben Whitelaw

I won't be drawn on my plans for the mouse. more on that in next week's podcast

Jen Weedon

Okay. Okay

Ben Whitelaw

Hello, and welcome to Ctrl Alt Speech, the podcast where we make sense of the major debates shaping online speech, platform power, content moderation, and the future of the internet. It's June the 18th, 2026, and this week we're talking about the UK's social media ban and the split screen reaction to it, where we are with the Anthropic Fable debacle, and for our Patreon subscribers, why Spotify haven't got their moderation act together and what that means for the future of the web. Ctrl Alt Speak is the weekly debrief between Mike Masnick, the founder and editor of Techdirt, and me, Ben Whitelaw, the founder and editor of Everything in Moderation. But as you've heard so far, Mike is traveling this week, and I'm delighted to have Jen Weedon as guest host this week. Jen, welcome to the podcast.

Jen Weedon

Hello. Thank you for having me

Ben Whitelaw

Jen, it's amazing to have you on the podcast. I've followed your work for, many years actually, and you were very kind enough to do an interview for "Everything in Moderation" back in 2022, which we'll talk a little bit about. you are-- I would call you a trust and safety veteran, which I hope is, not too pejorative a term. And you've worked at Meta for more than eight years. You've spent some time at the gaming company Niantic. Right now you're consulting and teaching at the Columbia School of International Public Affairs. you are the perfect person, I would say to, guest host the podcast. give me a kind of potted summary of your journey into the industry over the last decade or so.

Jen Weedon

Yeah. Before I do that, I wanna say on the veteran front, so, uh, many, many years ago, and this is related to my career trajectory, some friends and I put together this conference called CyberWarCon in DC, and this was, I don't know, maybe 10 years ago or something, and it was... The conference was sort of around, different facets of how governments misuse the internet to achieve certain long-term objectives. and one of the swag, situations we had one year was actually a CyberWar veteran patch on a, a hoodie. And so I've worn that hoodie over the years and have gotten like thank you for your service type comments, and I don't have the heart to be like, You know, I'm not, I'm not actually a CyberWar veteran," but I mean, maybe, maybe I am. Maybe I am. Um,

Ben Whitelaw

I think that's perfect. You should get, you should get more of those, badges made. I think more folks would, love to be recognized for their work keeping people safe on the internet.

Jen Weedon

that's true. That's true. Maybe that'll be my new, uh, my new Etsy shop.

Ben Whitelaw

yeah

Jen Weedon

yeah. So I, uh, just an abbreviated version of my career is, you know, I started off in cybersecurity and, and, in intelligence actually. but then parlayed that interest. I was always interested in the sort of human aspect of it, right? And rather than kind of putting my efforts towards protecting, computer networks, it was more like, what are the people, what happens to the people behind, these types of scenarios? And so that's what sort of prompted, my shift into trust and safety. so went, from a company called Mandiant, which did actually some really interesting work on exposing cyber operations, put on by governments. But then ended up working at Facebook, for quite a while, working in different types of harm areas. I would say, the longest effort was on kind of influence operations and surveillance, but I also worked on issues of hate speech, human exploitation, child safety. I sort of ran the gamut. but we were always focused on these topics from an emergent harms perspective, right? So, like what is the tip of the spear? Like, what are the worst threat actors doing? What are the most innovative approaches? and really sort of being able-- given the leeway to think very creatively and be very forward-leaning around looking around corners, and then trying to communicate kind of what are the risks that we need to be aware of. but also, and equally importantly, how do we root cause the vulnerabilities or failures that actors were able to exploit? And then how do we work with our teams to try to fix those? and eventually, I ended up, seeing that a lot and realizing that, some of the root causes of what we were seeing were things that were not easily solved and/or were chosen not to be solved for various reasons. Maybe they were difficult, maybe it conflicted with other important metrics. so then I started thinking about how to be more proactive. So that kind of sh- prompted my shift to work at Niantic, which is, going from a very specialized role at Meta to a more generalist role at Niantic, which had a very different perspective on trust and safety. Obviously, working with mobile games and spatial computing is a much different, trust and safety scenario than, a social media platform. and I still did some red teaming work, so but, you know, adversarial pressure testing of, new features, new products to try and think through how do we, how do we get this right before we release it, or at least how do we be better prepared. did that for a few years and then went to academia, which was, uh, an unexpected left curve, left turn. so teach trust and safety at the graduate school level, which has been wild.

Ben Whitelaw

Really interesting. I, I want to kind of come back to the academic left turn. what is it that about, you know, the last kind of few years has changed about that risk, intelligence piece? Like, I feel like the advent of generative AI has probably changed the way that teams like yours approach problems and issues on platforms, completely. have you seen that yourself as your current career has gone on and, and how do you think that will change?

Jen Weedon

I think the work of safety is changing dramatically. I think most of the foundational frameworks and taxonomies and, sort of ways of thinking about harm, are just being upended right now with AI and, particularly the way that, uh, and we can get more into this, purely the way that sort of internet is being restructured by AI as well as the major platform incentives and affordances. And so not only do safety workers in different forms need to recalibrate sort of how they think about what is a potential harm, but also where accountability lays with regard to that. and then also in doing the work, you know, there's increasingly, lots of vendors and, companies are, are AI first in their approaches, right? So using different tools or, LLMs for content moderation. And so the work itself is also being changed radically by AI in different forms. And so, on the one hand, it's great to have sort of a, bird's eye view of this being outside of, industry. On the other hand, I actually really miss being inside and, and sort of experiencing it firsthand. Um, but you know, I'm doing the next best thing, which is, talking to people and studying it. So, when it comes to teaching trust and safety, even our last syllabus last year, I think probably needs to be completely rehauled given how much everything has changed.

Ben Whitelaw

Yeah, it's not an easy task, teaching trust and safety at the moment. There's a kind of constant re-updating, the reading materials. for listeners who maybe aren't so familiar with, the way that colleges are thinking about trust and safety, give it, give us a sense of like why Columbia have started teaching this course and also why other universities like Stanford and like Cornell have really started to invest in producing programs that set people up for a career like you've had.

Jen Weedon

So I went to graduate school and graduated in 2008, and so I did Essentially, uh, I went to the Fletcher School, which is sort of Tufts' equivalent of the school that I teach at now at Columbia, which was, you know, around sort of international affairs, and we had a tech policy arm, but not really. In 2008 it was very different. I think, understandably, the last five years, I mean, I would say 10 years, but within the last five years or so, I think schools are further investing in, like, tech policy and, I mean, it depends on the type of the lens brought to this, right? So, like, obviously at an international relations and public policy school, maybe the emphasis is more on, policy and sort of geopolitical dynamics versus, an engineering school or computer science school, and versus, you know, there are law schools that teach aspects of this. So I think, you know, depending on the discipline, and the orientation of the school, the frame that they bring to this topic may differ. that said, I, I also believe it's really important to make sure that, folks coming up through computer science, you know, have access to, courses on ethics and whatnot, and likewise in the, international affairs, public policy schools, that they have exposure to some of the more technical aspects of how content moderation works, nuts and bolts, or how CSAM detection works, et cetera. so I think although even now, If ideally, you know, graduate schools are not just meant to, like, prepare people for jobs, right? They're, to learn things. Um, but, you know, we have to be realistic that, getting a job after is, is a natural sort of end goal that people want to have. But even now it's like the job market is, really, abysmal. and, and I think students now, you know, I've had many, many conversations of students who are like: Does this c-career even exist anymore?" Um, so it raises dual questions of like, one, what's the purpose of education, which is like an existential discussion, but two, What are the right skills to prepare for a safety career? and what does that mean, with how things are changing? And I think that the jury is still out on that. and I, I think there's a lot of smart people thinking this through and, trying to prepare, but a lot of it is the ground is shifting beneath our feet.

Ben Whitelaw

Yeah, no, it does seem like that. we won't go too much deeper on that 'cause I wanted to talk about a report that you've been working on, in the last few months, which actually really neatly coincides with your appearance on the podcast. It's coming out tomorrow, Friday the 19th of June. the report is called "AI Slop and the Information Ecosystem," and it's based upon a convening that you held in March this year of some experts and practitioners in the space. you kindly shared it in advance. It's a great read. Tell us why you decided to kind of focus on AI slop to begin with. is there more that we can say about AI slop?

Jen Weedon

There's a lot we can say about AI slop. I think, um, back in early at the beginning of this year, end of late last year, my colleague, Camille and I were chatting and, both of us are always, it's like, emergent, undefined areas of like consequence, but like we don't really fully know it, are catnip to both of us. and so we ended up, partnering with the Hewlett Foundation, to come up with the Slop Salon, which was w- really fun. Uh, we ended up, like I said, or like you said, convening about 20 folks across investigative journalism. We had Jason Kobler from 404 Media, folks from Frontier Model companies, academia research, people who like think about AI-generated content, particularly low quality, mass-produced, algorithmically amplified AI-generated content about what is the landscape of this topic, of this issue, right? And how can we think about it, and also what can we learn from the past? So the goal wasn't to like come up with a one-size-fits-all definition and solution. The goal was to get a bunch of people who think about this all the time to share their perspectives and then start to kind of map it out. and I think, you know, we can get into the definitions in a second, but for me also in designing the workshop, being both in academia as well as coming from platform, I was very mindful that often when you get platform folks in a room, they seek to kind of bound the problem and think of it for like how is this gonna be operationalized at scale? How do we kind of frame it? Because that is a natural instinct, right? That is literally how you have to manage these things. And yet the concerns of jumping to conclusions or, putting into solutions without understanding the problem fully is something I wanted to avoid. and that tension came up. But, we ultimately kind of, even though we had a lot of disagreements on like the definitions, the scope, et cetera, we, we all agreed that this is not a content moderation story. slop isn't just, you know, it can be defined, and the most common definition is really that it's the low quality kind of mass distributed, algorithmically amplified, AI-generated content. but that's kind of like where the agreement ended. I think there was, so many different perspectives on what are the harms? What are the impacts? How do we think about this? What's the scope? What are the properties of slop? Et cetera. So I encourage people to read the report.

Ben Whitelaw

Yeah, I mean, definitely go and read it. It's, it'll be in part of our show notes as usual. I mean, I really liked in particular, Jen, the way that you kind of, AI slop as a term, as a phrase to fake news, and said, you know, fake news as, as a term has kind of lost all meaning. It's become collapsed into, uh, kind of various different ways. It's being used by different groups to mean different things and, it's almost meaningless now, and that there's a risk of AI slop kind of going down the same route. can you talk a bit about, like, kind of the discussions around that very basic idea of defining what it is?

Jen Weedon

Yeah. So, on that front, I think, we tried to balance both, leaning in a little bit to the sort of platform approach of like, okay, well, we need to bound the problem somewhere, right? So we came up with this notion of the slop funnel, where we kind of had a, a spectrum of, have AI-generated content is like the mass category, right? Within that, there's things that are kind of like more clearly defined as slop versus less. So like on the, far left of the most widely distributed and sort of least directly harmful is, AI-generated content that maybe isn't, high quality, but, maybe it's stuff that like if you're working and your colleague sends you a draft of something that's, you know, been done by Claude or ChatGPT and you're like, "Ugh, okay, well this," you know, you can tell that it's created by a chatbot, and then you have to spend extra time. Like you know, work slop, right? and that has... There's some articles on that and whatnot. then you kind of get into like, all right, most, ambiguous area, which is this sort of lots of the cats jumping on trampolines, shrimp Jesus. You know, I, I had a, polar bears with barnacles on, on them kind of spate of slop in my feed for whatever reason. You know, the things that your, far-flung family members are likely to send you to be like, "Look at this," and you're like, "Oh, okay, that's clearly AI." Like there's a big bucket there where you're like, we don't really know the volume. We don't know what content this is displacing. we don't really know what the harm is yet, right? Short-term or long-term. And then we kind of, then there's this next category down the funnel, which is AI-generated content that the harms are a little bit clearer, but there's still a lot of ambiguity. And so this would be like your slopaganda. So this is, we can get into what slopaganda is, right? If you remember a couple months ago, the Iranian Lego slop or maybe the, um Actually, I wouldn't characterize that as slop, as an aside. or, you know, your White House, videos with, AI-generated, defecating on protesters ki- type of information. and then there's the AI-generated content that is not slop at all, but that is also harmful, which is like your NCII or your CSAM or whatnot. So we kind of mapped out like where should we focus, and it was that kind of middle, those middle two chunks of your cats on trampolines, your, arctic animals with creepy barnacles of like, okay, where's the harm here? And then the, the kind of questionable, politically themed, maybe misinformation and there's some great examples I can point to. those were the areas that we all felt were like worthy of more, intentional research.

Ben Whitelaw

Interesting. And not least because platforms are, making that content very easy for users to, to create, right? You know, it's-- if you go onto a Meta platform now, whether it's Facebook or Threads, you are being kinda shown that y- it's very easy to create, content of a, young woman who looks like she's wearing a football kit but is also in a battlefield. You know, this kind of like know, obscure, almost like, anything is your imagination kind of content, which as you say, we don't know the impact of, in terms of our attention, in terms of the kind of downstream effects. the report talks a bit about that. We probably w-wanna get to today's stories then, but is there anything else you wanna note to listeners about what the report tries to do or what the benefits are of reading it?

Jen Weedon

I mean, I think the, the story of Slop is really a story of like how are platform incentives, monetization schemes, and, and this sort of like new content economy changing the structure of the internet? And like one specific example that I think that I can wrap this up is, is, you know, there's a, a story in there of a 17-year-old content creator from Burkina Faso who made a fake AI video of a coup in France that got, I don't know, 12 million views. It got President Macron to say like, "This is not a real coup," and asked Meta to remove it, which they did not, and the creator made like seven euros off of this. But, for us it was, it was sort of emblematic of if you can easily... 404 Media has done some great reporting on this, as have others. But if you can-- If there's this sort of strengthening ecosystem and machinery of content creation of people who are not... Who are just trying to make money, and they can go easily find lots of lessons on how to make, content and optimize it for various algorithmic distribution or reach. if there are topics that are political that get traction, you know, who needs troll farms? You know, you have this like machinery that's like very easily, setting folks up for all manner of, content distribution of different types. But what does that mean for, you know, how... this content continues to saturate everything, what does that mean for our own ability to discern, truth from fiction? What does it mean for platforms who are trying to figure out rules of the road with content that's AI generated that maybe wasn't done via coordinated and authentic behavior, was done by someone who's just trying to make a quick buck, but who happened to align on political co- So it's just, for me, this report kind of teases out what really is, I think, a fairly profound change, that we are not super equipped for, in our opinion.

Ben Whitelaw

No, I think that's an understatement. And, uh, yeah, there's, there's lots of reasons to read the report. It's a great read, and it's a great bunch of people that you convened to have that discussion. So worth our while. before we jump into today's stories, Jen, I just wanna thank our growing group of Patreon subscribers. Um, we've got our supporters who pay $10 a month to get extended episodes, additional analysis, and also early episodes. We, we share our episode with our Patreon subscribers before we do our free one. So if you like listening, kind of late, Thursday UK time, then you know, you wanna become a supporter. Founders and insiders also get all of the above, plus the chance to recommend stories to us to cover, and that's what keeps the lights on, on Controllable Speech. so yeah, Jen, thanks for kind of outlining the report and talking to us about that. we have what you called, a, a whole gamut of stories to get through. are our spreadsheet this week was kind of chock-a-block. we're gonna start with a story that I think many people will have been paying attention to, which is the UK social media ban this week. or rather the kind of outline of a social media ban, which I think is fair to say. there's not a lot of detail on this, but we will talk about it nonetheless. Um, if people have not been following it, Sir Keir Starmer, the UK Prime Minister, got up on Monday, gave a press conference in which he outlined, the under-16 social media ban in the UK, which would take place, it would become implemented in spring 2027, and it will, apply to a bunch of platforms, many of whom we know. TikTok, Snapchat, Instagram, Facebook, X, and YouTube. Not YouTube Kids. so, so only the kind of major platforms. Unclear exactly what the, reasoning for those platforms is. many people pointed out that there were no gaming companies included in there, and Starmer was very, clear to point out, probably because most politicians in Westminster use it, WhatsApp, and Signal, were not included. So a focus on peer-to-peer type social media platforms that will no longer be accessible by, kids under 16. There was a few other details that, again, were, dangled but not really discussed. A few kind of product features, Jen, that, that he mentioned which won't be available to, 16 and 17-year-olds. so live streaming and also messaging people that you don't know on platforms. It was kind of termed messaging strangers. I think that's designed to appeal to parents. It's probably focused on, DMs from people who you don't follow or don't know to combat an issue that we all know happens where children are, are approached by, by spammers and scammers and people, who are, trying to cause harm to children. But that was, a kind of fairly large part of, the announcement without really going into detail. And there was another element, was again, mooted but not really, discussed in detail, which is a digital curfew. Starmer said that there would be a discussion around an 8:00 PM, I think it was, digital curfew for children but that more information was gonna be discussed in July. and the reasoning, for all of this was kind of reducing harm for children, improving their wellbeing, ensuring that they have a healthier childhood. And we could talk a bit about how that language resonates quite strongly, Jen, with some of the campaigning groups who've been such a force behind getting UK politicians to, take this seriously. I wanna talk about the reaction primarily, Jen, because there's not a lot of policy to go on here. it's all fairly light. But what did you see about the ban from where you're based in the States? What was the reaction? What kind of coverage did you read about it?

Jen Weedon

I actually haven't, I haven't read a lot of reactions to the ban other than, you know, tech insiders and my own echo chamber of people who focus on this. I do wanna share, know, when we were preparing for this episode, you mentioned the sort of differing reactions of like folks who are professionals, and experts in this space in different ways, and then parents and these sort of like, these groups kind of having different narratives. And so I am both a veteran of this space as well as a parent. Uh, my child's eight. I'm also an elected school board member, which means in the US it's the, the sort of elected governance board over public school districts. which, the reason this is relevant is that it, you know, has to do with tech device policies in school and curricula and things like that. Um, and even in my state, in Massachusetts, I think they recently, passed a law social media ban, although I can't-- I don't remember the exact details. but for me as a parent, and maybe this is Delusional. but the, the death of nuance and, people's inability to sort of converse on like First of all, where are kids in this, situation, right? Like you know, I've sp-spent a lot of time recently talking with folks on sort of children's rights and, and a rights-centered approach towards this. and I, can imagine that, the perspectives of youth and teenagers are important in this. also because their perspectives also relate to how they adapt to these bans, right? If they're using workarounds like VPNs, if they're going to other spaces, or if they're just as-easily able to access things in ways that, adults may not understand. so the sort of like who's at the table and, and what kind of perspectives are we privileging is something that I think is relevant. the death of nuance regarding like, the commentary on like why aren't games, you know, are we talking about what types of games? Are we talking about games that have, UGC in them, messaging, Minecraft? Like these are all very different games that kids use in different ways. and this sort of narrative that like all games are terrible or all social media is leading to X is just so, I think, uh, such a useless way of thinking about it. but I, I also think this relates to like how do you-- what are better ways to engage the public on this? Because clearly like experts who talk about nuance are not communicating effectively, or are being outspent by, lobbyists and whatnot. So, I also think that, removing agency for children and then also having more sort of surveillance mechanisms is, is not gonna be long-term helpful for anyone, right? I was reading some of the expert commentary on this in the UK, and an expert said something that I, the phrase I really liked, which is like, we need to create the conditions for competent and critical engagement with tech versus, prescribing, a one-size-fits-all, because I think there's just so many anticipated consequences that I can see that are, are not helpful for children. and it also doesn't necessarily, hold the, tech accountable for their design decisions. it also sometimes lets parents and schools off the hook for putting in the work with their kids to help them learn. You know, I was, I was literally talking to my son this morning my eight-year-old, about chatbots he has access to at school. and also how to interpret like summaries when he uses, Chrome or Safari and the differences between using that summary versus LingQ. So, uh, I have a lot of thoughts on this and I, I think it really misses the mark on so many different levels.

Ben Whitelaw

it's, it's funny, isn't it? Because that was the consensus among, policy experts and trust and safety professionals in the aftermath of this announcement. You, you know, it was kind of ubiquitous reaction that this was, a headline-winning proposal, but not something that was gonna be very easy to enforce or that would necessarily have the right consequences. we know that the Australia ban, which is the kind of earliest of these social media bans that came into place last December, has not got strong evidence, for, addressing the root cause that many people are concerned about. And yet, the UK government has gone ahead. It has of sped run the consultation at the start of the year, got 116,000 responses to it, but as you say, only a- about, I think 13 or 14,000 of those were from children, so probably around 10% of, the overall consultation, which backs up your point about, are they, being listened to and are they at the table? again, we don't have a lot to go on from, politicians, but the media reaction I found very interesting. You know, you had a lot of focus actually on children. There was a lot of people, a lot of outlets going and sticking a microphone in front of children, and I, I found it super interesting, Jen, that how clear people's views were on a ban. You know, they weren't all in agreement, but they had really strong justifications for it, which made me think, actually we, we've missed an opportunity here to bring them into the policy discussion, over the course of the consultation. You know, some kids were saying, you know, "Parents should be responsible for having the conversations that you have with your son." and others were saying, "Look, some parents just don't have the capacity. They don't have the head space, they don't have the time, they don't have the skills, they haven't worked in the industry like you have. and so maybe that's not, you know, we should create a kind of baseline, element of safety here." and then, you know, there was the kind of parental reaction. we've talked a lot about the Smartphone Free Childhood campaign, which, in the UK has been incredibly successful at driving people to the consultation and to organize, and many of those regional leaders from that campaign group were actually at Downing Street for the announcement. So it's very clear that the, Labor government have, recognized the concerns of that group and of parents generally, and, you know, acted upon them without necessarily, thinking more broadly about what happens down the line, or what, what children think. and as I mentioned before, this is partly because, Labor, as a UK government are under pressure right now. There is a, a by-election happening in a, a northern town in the UK that has significant ramifications for S- Sir, you know, Sir Keir Starmer and what happens next, you know. Imagine if, i- There was a vote happening in the US that might see Donald Trump lose his, presidential seat. You know, that would, be a big, big news. So that's what's happening in the UK. It's led to this ban, and there is this kind of split screen reaction. Policy experts who, who are slightly wringing their hands and, worried about how this will play out, trying to shape the next steps, and then also the parents who are, are frankly, jubilant, that finally they're being heard and, and seen and kind of listened to in a way that, they feel like they haven't been before. you talked to one of the-- about the age assurance element, Jen, which is, you know, a massive, massive conversation. There are states that are bringing in age assurance and age verification technology in the States. That was a kind of like, afterthought in the announcement. more information is gonna be set out by Ofcom, the UK regulator, in due course about what platforms will need to do as of, spring next year. What are your views, about that kind of age assurance piece? Like, are you as concerned as others about that might go?

Jen Weedon

Yeah. I mean, I, I think the, the, like, meta concern here is that, you know, there's all these initiatives, but the enforcement realities are too often an afterthought, right? It's like sort of magical thinking. the Anthropic story, which we'll talk about too, is like, you know, there's these sort of dictates of like, yes, remove access for all foreign nationals, and you're like, "Okay, how does, how does that actually work?" Like, how... You know, so there's this lack of understanding of one, sort of how different technologies work. Two, how, means of circumvention, and, side effects are in terms of like norms of surveillance or like how it disproportionately affects different populations. but none of that seems to be like part of the actual discussion earlier on. so I, you know, again, I get back to like what part is most broken here? I also, I was interested that you mentioned that YouTube Kids was a, carve-out for this, which is interesting because if this is all ostensibly about protecting kids, YouTube is, YouTube Kids in particular, this came up in the slop work, has a prolific amount of AI-generated content that has absolutely no educational value, right? That it, it is just served up without searching, right? So it is algorithmically put in kids' faces. there's often like no narrative or educational value to these things, but it's so visually appealing and rapidly changing that like sort of the dopamine receptors go bonkers. And so it's like, okay, so on the one hand we're saying, "Yes, we want to protect the kids. We want to do all these things." On the other hand, we have YouTube Kids with all sorts of, slop and, and sort of problematic materials. at least in the US, we have kids on devices for much of their day, right? with access to a lot of the same, services that often are purportedly wanting to be blocked or the kids will find workarounds. So there's just so much, incongruence or sort of it just, it's very hard to make sense out of

Ben Whitelaw

No, uh, totally. And I, I, I wrote this piece for "Everything in Moderation" in which I, wrote that I just felt disorientated by watching this whole reaction play out. You know, the, the incongruence is a great word for it. It's like, you know, in some ways I always felt like I was being gaslit by by like one of these groups. It's like, neither group is seeing it from the other's perspective, and we don't have a kind of shared understanding of what the problem is, and so we've gone off in, different directions. and it, and it ironically, to bring UK politics, of 2016 into it, but it felt very Brexit-like actually. Brexit was a big moment in the UK. It was one group of people who didn't feel heard about the way that the country was moving and, kind of reacted against the, politicians and the powers that be at that time. And there was a group of people who said, you know, from the process perspective and from the way that we're kind of integrated into this wider, geographical block, we can't move. It's impossible. We... The benefits are too small. When it, uh, there's parallels there, I think, with the social media ban. the kind of technocrats versus, these kind of emotionally, driven voters essentially. So, again, I, respect the campaigners for managing to, shift public opinion and, and actually, like, move people to, input into consultation and to make something happen because that is increasingly difficult in today's political landscape. But, we don't know where this is gonna go, and I feel like we will be returning to it a lot on the podcast, uh, which is both good and frustrating

Jen Weedon

I mean, you said one thing that actually stuck with me, which is, the privilege that some of the perspectives I have, like, you know, that I, you know, I've worked on these topics and somewhat, fluent in some of them, and I also have the time to talk with my kid about these things. And I mean, I, I often don't have time and I make time, but, that is not a, an expectation that is universal. And so, this sort of perspective of well, yes, parents should do more, like there's certainly something to be said and something for my, my own self to kind of reflect upon, which is like, how, to what extent is my own perspective, do I need to put it in check also with the realities of many, many other people? so thank you for that.

Ben Whitelaw

not intended, but, um, yeah, I think it's one we all can think more about. we'll put that story to bed, Jim. We'll talk a bit about, a story you mentioned there, the Anthropic fable debacle, I'm gonna call it. and another story from India that I think come under what Mike, if he was here, would call, examples of the despotification of the internet. he talked on the podcast a few weeks ago about an essay in which he kind of outlined how governments, when incentivized incorrectly, and given, access to particular choke points on the internet, they will do things that run counter to citizens and, users', best interests. And I think these two examples of stories are, perfect, examples of that. talk us through, again, a lot has happened about Anthropic since last Friday, but give us a kind of overview of where we're at with that story.

Jen Weedon

So my understanding, and I, I think, again, like the, news and the facts keep dribbling out, so all this is subject to change. But, you know, my understanding is that on early last week, Anthropic launched Fable, one of its Mythos-related models, for general use, but with specific guardrails. And on Friday, on the 12th, Anthropic shut down access to both Fable and Mythos after a dispute with the Trump White House, apparently around, jailbreaks identified, in Fable that the White House didn't necessarily share a bunch of information for or believed that they were more consequential than, than they were, and implored Anthropic to, for national security reasons, limit access to this model for foreign nationals, um, which gets to the enforcement question. So there's, there's lots to unpack there. brings to the enforcement question was Anthropics like, "How do we, how do we actually do this?" and the result was to, pull it off the market because of this, demand from the, the White House. and essentially, I think this raises, a number of questions, but for me, one of the takeaways was, one, this seems to be yet another example of, the US administration's sort of incoherent and whiplashy perspective as it relates to AI governance. governance writ large, but particularly around AI, uh, safety and evaluations practices. the second part being what information should be shared justifying why these decisions get made, right? with Anthropic kind of reporting that they, felt like they had a good understanding, that actually some of the jailbreaks that were identified that, you know, other models were able to find, so this wasn't something that would, lead to widespread sort of ability to create, quote, "cyber weapons," which for me, I don't even know what that means. It's like a really imprecise term. and also, y- on the flip side, like there are ongoing discussions about the need for third-party evaluations around, you know, particularly in certain sectors or types of risk, right? but even with that, there's discussions around, which areas are rightfully, the most concerning or high risk versus like what are we willing to accept, as course of business, right? Like, are, are people having mental health catastrophes or not being able to discern truth from fiction, like that's cool, but we don't want cyber weapons. You know, so there's a lot of kind of embedded in terms of uh, how we make decisions, and I think last thing I'll say is there's been a lot of commentary around, Anthropics own messaging and positioning itself in the market, and how, a lot of how Anthropics sort of long-termism and ways of thinking are, are setting the stage for a lot of, either academic capture or other, you know, marketing. Like there's just, there's so many ways in which that narrative kind of takes hold, and I think, Yeah, there's many threads we could pull, but that's my understanding of the Anthropic debacle over the last week.

Ben Whitelaw

Yeah, I think you've done a really good job of, of summarizing that. You know, the, the, long and short of it is that the Fable 5 model is still unavailable to people, outside of the US, and in the US as well, I think.

Jen Weedon

Yeah. No, it's not available in the US

Ben Whitelaw

uh, so whenever you go onto Claude, it's like, "You cannot use this model." and it seems like nothing has really moved. I haven't seen any reporting in the last couple of days that suggests that's going to change any time soon. but as you say, the reasons why that's happened, continue to kind of become a bit clearer. One thing I, I will note is that there is a bit of a backlash that's been emerging of cybersecurity experts and, people working in, AI model development to, get the US administration to do exactly what you mentioned, which is to be more coherent, to be transparent, and to lay out a plan for how models will be, evaluated and, standards will be assessed. Over 100 people have signed a letter, which is available at freefable.org, which lays out, you know, exactly why the administration should, come to the table and lead in terms of establishing a process for evaluating models. And, it plays to the administration's kind of national security risks, its concerns about falling behind in the AI race. It's a very good letter. Alex Stamos, who, uh, many of our listeners will know, he's been involved in it as well. So it's a, clear-sighted call for, a bit of clarity. And, again, I think no one wants to, to be in a world where a, government of any kind can call a model back, and have it retracted from users in the way that, we've seen over the last week. It's not a particularly good look for anyone. so yeah, Jane, great job of, of summarizing that. I don't think I would be able to do, anywhere near as good a job. we're gonna share another story now, which I think speaks to these topics, very acutely, and that's how in India, Telegram has been at the center of a story, as a result of being the kind of hub of medical exam fraud. so I was, I was kind of reading up on this, this week, Jen, but there's a massive exam that gets taken in India called the NEET test, the National Eligibility, Common Entrance Test, and it's a, a big one. Anybody who wants to go to medical college in India has to take it. And it was taken in May, only for people, for the 2.8 million people who took it to find out that the, the results were scrapped because they found out that some of the, papers and the results had been leaked in advance via Telegram. The exam is being retaken next week on the 23rd of June, and so what the government has done is ban Telegram for, a series of days up until the exam has been rolled out. And that's clearly to kind of prevent a repeat of what happened before. you know, you can't even imagine what it must have taken to set up exams for 2.8 million people across a country as large as India, and they clearly, don't want a repeat of that. But anybody who wants to use Telegram for anything else apart from leaking exam papers will not have access, which is a kind of shocking turn of events, and, and Reuters explains how, you know, many average Telegram users have been caught up in this. Again, this, this feels along a similar line, Jen, doesn't it? You know, a government taking a universal stance, without any justification, and kind of claiming, national security or, you know, citizen safety.

Jen Weedon

Yeah. as you were speaking, I actually just had a memory from, some of my early days at Facebook when I was... One of the tasks of our team was to work with the policy team to monitor outages, right? there was, I remember, this was probably 2014, I mean, there have been many, periods of outages or governments, restricting access to Facebook at various times for various reasons. But one of the instances, I think there was, uh, I can't remember which country it was, but, uh, there were cheating allegations of like... I, I don't remember exactly what happened, but it was, the same thing where we're like, what the heck is happening?" And then sort of digging into it and, there was something around, widespread cheating and the, the, best solution that they could come up with is to restrict access to a, a major technology for the time. So it's, you know, I, I would love to be inside the heads of the policymakers who are like, "Yes, this is the best solution," is to just try and, take this massive bludgeon to like people's access to, messaging apps. but it, seems to be a repeat tool of statecraft, and I'm not sure it's gonna go away anytime soon.

Ben Whitelaw

No. And, as we were talking about the Anthropic story, I, I actually don't think there's a, an incentive for governments, elected officials to come up with a better way of doing things. that's the, the crux of it, isn't it? You know, there's no real rush for the US administration to come up with a better way to evaluate models because they can just issue directives to private companies, and say that model cannot be used for various reasons of national security. You talked about incentives from an AI slot perspective. That doesn't seem like a particularly strong incentive for, administration officials in the US or the Indian government to do anything differently because they have, power right now. how do we change that? how does that change, do we think?

Jen Weedon

I mean, Ben, if I knew the answer to that question, um, you know, I, I wouldn't be sitting here right now. Uh, you know, I don't, I don't know. We haven't... Uh, it hasn't seemed to, hasn't seemed to have changed, in my lifetime. It still seems to be this sort of like, saber-rattling, with people in power, you know, I don't, I don't know is the answer to your question. I don't. I

Ben Whitelaw

No. No. W- w- I put you on the spot there and, uh, 'cause I also don't have an answer. but both stories I think speak to the same theme and that's why they're, in our, small story section. we're gonna kind of wrap things up there for our free listeners now, Jen. we've talked quite enough about social media bans and, the ability of governments to, do whatever they want. we're gonna talk more about a really intriguing story from Spotify and the way it moderates some of its very, very large five million podcasts, on its platform. What is it about this story that really caught your eye before we dive into it?

Jen Weedon

So this story, I think originated last year with a state rep from a state north of me, so I'm in Massachusetts, with an inquiry to Spotify about podcasts that had information on how to access, or purchase illicit drugs like fentanyl or, different types of benzo- diazepines. And so the type of content was podcasts that, didn't get a lot of views, but the investigation found that indeed this content did exist, but it didn't really get any views. and so the questions, as a result are like, is this a content moderation issue? Or if a fentanyl podcast, is posted on Spotify and nobody listens to it, does anybody-- is there any point? and so I think that there's some stuff to unpack,

Ben Whitelaw

Brilliant.

Announcer

That's the end of this week's free edition of Ctrl Alt Speech. To hear the extended episode with additional coverage, become a Patreon supporter at patreon.com/ctrlaltspeech. That's C-T-R-L Alt Speech. If your organization is interested in sponsoring the podcast, contact us via our website at ctrlaltspeech.com. Thanks for listening

Contributor

Ben Whitelaw

Host
Contributor

Mike Masnick

Host