Ctrl-Alt-Speech
Ctrl-Alt-Speech is a weekly news podcast co-created by Techdirt’s Mike Masnick and Everything in Moderation’s Ben Whitelaw. Each episode looks at the latest news in online speech, covering issues regarding trust & safety, content moderation, regulation, court rulings, new services & technology, and more.
The podcast regularly features expert guests with experience in the trust & safety/online speech worlds, discussing the ins and outs of the news that week and what it may mean for the industry. Each episode takes a deep dive into one or two key stories, and includes a quicker roundup of other important news. It's a must-listen for trust & safety professionals, and anyone interested in issues surrounding online speech.
Follow us on Instagram, YouTube, and Bluesky.
If your company or organization is interested in sponsoring Ctrl-Alt-Speech and joining us for a sponsored interview, visit ctrlaltspeech.com for more information.
Ctrl-Alt-Speech
Make America Break Again
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
In this week's episode, Mike and Ben cover:
- Widespread outages hit banks, air travel, the NSA and Apple, disrupting seemingly everything (CNBC)
- National security leaders say recent outages are unrelated. Washington remains skeptical (Politico)
- Our Dipshit DHS Secretary’s Outage Expert: “One Guy That’s Very Smart With Tech” (Techdirt)
- At Nasdaq opening bell, Anthropic CEO recommits to pacing AI development (TechCrunch)
- An Open Letter from Members of Teen Online Safety, Wellbeing, and Youth Councils Around the World (Snap)
- The Social Media Ban Study (University College London)
- How I accidentally became a millionaire (Lyra.horse)
And in the extended episode for Patreon supporters, they cover:
- Strengthening the fight against online child sexual abuse with new tools and support (OpenAI)
- Platforms must take stronger action to combat child sexual abuse material online, under new Ofcom rules (Ofcom)
- Reports of AI-related child sexual abuse material rise 8,400% in two years, data shows (The Guardian)
Our fun links this week include Nerdticker and Internet History Roulette.
If you’re already a Patreon supporter, you can get the extended episode on Patreon.
Follow us on Instagram, YouTube, and Bluesky.
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).
It's good to have you back on the podcast, Mike. It's been a few weeks, and as such, I expect you have a number of different fun things from across the internet to share. that's probably all you do when you're not on the podcast, right?
Mike MasnickThat's right. I'm just, just looking for fun things on the internet. That is my entire life
Ben Whitelawso what's the latest on that list?
Mike MasnickUh, so the latest one, I actually just saw this last night. It's by, Louis Gray, who, I don't know if you know him, he's one of the original tech bloggers, like from way back in the day. He works at, Google as like a product guy But he's been messing around and building some really cool stuff lately. And so he released yesterday this thing which you might find useful as well, and I think many of our listeners might find it useful. It is called Nerd Ticker, nerdticker.com.
Ben WhitelawHmm.
Mike MasnickAnd he had just built an RSS reader, and he has bolted onto it, you can ring the bell. it will highlight interesting nerd-related tech stories with Blue Sky discussions on them.
Ben WhitelawUh,
Mike Masnickand so it's a nice little aggregator the design is great. It looks really cool, and you can follow important tech news with it
Ben WhitelawThat's fun. I mean, I don't have my Bluesky bell, but, um, it, somehow disappeared. But yeah, it's, uh, it's worth ringing. you are on the Bluesky board. But that doesn't mean that this isn't a cool, fun thing. I'm definitely gonna check this out. This sounds great
Mike Masnickyeah, I mean, like for something that was just launched yesterday by one guy as a sort of side project, partially vibe coded, I believe, that like to have it look this polished is, pretty impressive, so
Ben WhitelawThat's super fun. That's super fun. my kind of fun thing is less polished, I'd say. It's a, it's a little bit kind of, old school in its aesthetics, but it's, a website called internethistoryroulette.com, which I think you'll like. I discovered it via, a good kind of friend of the podcast, the guys at New Public, they shared it in their newsletter a few weeks ago, and I've kind of used it occasionally to pass the time. it's exactly what it says on the tin. You can spin the roulette. You get this website, often very kind of old school, website served up from the Internet Archive's Wayback Machine, and you get to kind of see it in all its glory. It has some fun annotations about, you know, the World of Warcraft, website from back in the day, or collegehumor.com, which I think is still going. And it's a nice way to kind of explore the relics of, the internet as was. So, you know, good way to pass the time, when there are so many outages on the internet as we are gonna talk about today 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 November the 19th, 2026, and this week we're talking about who's behind the major digital outages, safety being at the center of Anthropic's IPO, and for our extended Patreon episode, what OpenAI and Ofcom announcements mean for child safety. Ctrl Alt Speech is the incredibly investable weekly podcast by me, Ben Whitelaw, founder and editor of Everything in Moderation, and Mike Masnick, founder and editor of Techdirt. Mike, you're back, after seemingly, uh, the most flights that one man can take in one week.
Mike MasnickYeah. I mean, obviously I've been traveling a lot. last week I missed the podcast because I was in Miami for a conference where I gave a very fun and, enjoyable keynote speech, and, and then I was all excited to fly home finally after a few months where I've been traveling quite a bit, And, my flight got delayed, because of weather, and then literally as I was boarding, there was the now infamous FAA ground stop of all flights in US airspace. and I, then spent five days trying to get back home. Uh, and I finally got home two days ago, five days after I had intended to be home, and with my family. So it has been a very exhausting week of, just, you know, trying to figure out ways to get home. I had multiple flights canceled. I was supposed to travel through all different cities to get home. Eventually I did fly home through Dallas, and did make it home late on Tuesday night. But it has been quite a week for me
Ben WhitelawYeah. Yeah. quite the journey. Quite the journey. I mean, it's gonna be a, a story we talk about today. the FAA outage was just one of a bunch of outages, and I think it speaks to, the kind of fragileness of our digital infrastructure, which I think is, is something we wanna talk about on the podcast. before we do, I think it's worth just flagging to our listeners that there isn't gonna be an episode next week here on Ctrl Alt Speech. you're gonna be celebrating Thanksgiving, and I'm away as well. So we don't have a special episode at all. but you can go back and listen to the recent Social Reckoning episode that we did, which has got some really great reviews and comments from some of our listeners. People really liked us delving into the world of film criticism. Uh, we, we tried our best. And, uh, yeah, if, you haven't already become a Patreon supporter, now's the time to do so. you can do that by going to patreon.com/ctrlaltspeech. Gets you extended episodes of the podcast every week, and also earlier episodes, so you can listen way before the weekend rolls in, and, uh, that's worthwhile. also worthwhile is rating and review us wherever you get your podcasts, even if you're not ready to stump up your hard-earned cash, and we know it's tricky for people at the moment in, lots of places around the world. it's free and it's simple to give us a rating or review wherever you get your podcasts. Big help to us getting, discovered.
Mike Masnickrecommend us to your friends and coworkers and everyone you know because you know that they would learn a lot by listening to us. So share, share the podcast
Ben Whitelawthat might involve, you know, going to one of our newly minted, social accounts where we have now put out clips of our videos and sending those to your family WhatsApp, to your colleague that you don't despise, and, uh,
Mike MasnickOr maybe the ones you do despise. Who knows?
Ben WhitelawAll of the above. so let's, let's crack on then, Mike. We, are gonna start with, that big kinda digital outage story this week. as well as the FAA one that you mentioned, there were, were plenty others that took place in the space of just a few days. Microsoft, JP Morgan, Apple, United Airlines, you know, they kind of, every day there was a essentially a new news story about, some part of our digital infrastructure kind of collapsing, in real time it seems. you were at the heart of it in terms of y- you know, you struggled to get home. Um, but I wonder if you can unpack and make sense of these outages and, and how we should be thinking about them, 'cause I know you've got a lot of, questions from, Techdirt readers and friends saying, "How should we be thinking about this? What does it all mean?"
Mike MasnickYeah. I mean, I think everyone was looking at, you know, the first big one was obviously the Microsoft one, which went down last Thursday. and then, the FAA the next day, and then the banks, and then United, then Apple. All of these things, I think, happening in rapid succession and often, you know, having consequential, results caused a lot of people to wonder. And the questions that kept coming up were specifically like, are these cyberattacks? Is this some sort of, act of war? You know, a lot of people were saying, like, Iran was trying to attack US infrastructure or something like that. Or the other big assumption was, is this AI, right? You know, we've had all these concerns about AI and AI going rogue and all of these things. and then on top of that, you know, with the releases earlier this year of the more cybersecurity-focused AI models, we know that different vulnerabilities have been discovered and there's been this rush to patch things. So there have been all these questions about what is actually happening and could it possibly be a coincidence, or is this something bigger and should we expect worse? And so I've been talking to a bunch of people and sort of exploring all of the available research and sort of trying to understand what is actually happening. And the simple fact is, like, we don't fully know. And so a lot of this is, pure speculation, but somewhat informed speculation I, I think. Um, there doesn't seem to be a clear indication of the two biggest things that people were talking about, which is that this was a big cyberattack, that somebody's deliberately trying to take down stuff, or that it is AI systems, beginning the, enslavement of humanity or, or whatever the, the goal is. it doesn't appear to be either of those things directly. And so then there's the question of like, why is it all happening at once? And there's, there are a few different theories there. the one is that, you know, humans are pattern-matching animals and sometimes, a bunch of stuff just happens to happen right in a row and we assume that they're connected when they're not. so that may be part of it, right? when that happens, you tend to like pay more attention. You know, things like the banks going down or United going down and even Apple going down, like those were all fairly brief outages and arguably in any other week if any of those happened, it would be a passing story and people would forget about it and wouldn't connect them. The fact that it happened after these other big outages makes it seem like a bigger deal. So I think there is some of that to it. The larger kind of working theory that, I find to be the most compelling in terms of what is actually happening is related to AI, but not as directly as most people are making it out to be. And that is that a lot of these bigger companies, including Microsoft, including the banks, probably including United, I, I don't know if I've seen anything official on that, have, either greatly diminished hiring on the tech side or, engaged in serious layoffs in the belief that AI tools are now allowing them to do more with less, which may or may not be true. I tend to think that, and this is a broad statement, a broad generalization, that companies that are laying people off and believing they can replace them with, AI are making a huge mistake and making a really bad bet. I think AI is a very useful tool in combination with a human, that they allow humans to do more and produce more where the human is in control. I think the companies that are sort of rushing to get rid of people are making a mistake, and this may be one example of that, where when you are rushing out the door, people who have worked in these places for a long time, with them goes a lot of tacit knowledge, knowledge that is not explicitly understood by the company. It's not within the systems. It's not within the documentation But there are often, IT people or techs, you know, engineers in the company who actually understand the little things that make things work and the little things that will make things break that are not clearly documented and people don't understand, and those people are being shown the door. And so when you suddenly do have to do an upgrade or a patch for something like that, and some of those patches may be because of vulnerabilities that were discovered by AI, you don't have the people around who would notice very quickly, like, "Wait, wait, wait. Before we do this, we have to make sure we, you know, put in this level of redundancy," or, "We need to set up this," or, "If we patch it, we have to patch it this way so that these other cascading failures don't happen." A lot of those people, I think, have been pushed out recently, and so a lot of the failures are when we lose that kind of tacit knowledge, the, I refer to them as sort of the load-bearing employees who make everything work and really understand the complexities of, how to keep these systems running because they're very, very complex, um, and very, very confusing. And, there are always some people within these companies who really understand them, and when those people leave, then these, types of problems tend to happen more often, where you have these sort of cascade failures because nobody quite understands how the different systems or the different tools or services interplay with each other. And so if I had to, put a bet down on like what, what is actually causing these to happen, it is the end result of losing a bunch of these employees by companies being short-sighted in their, hiring and firing practices. And then that leads to, just making the systems themselves more fragile.
Ben WhitelawYeah, I mean, it, it kind of makes sense, right? Because the FAA one in particular was, I think as a result of a third-party contractor who kind of made a change, right? And, uh, it's often the case that there are humans who interface with those third parties or who kind of, again, have that underlying knowledge about what has to change when they make a change. And it does, feel from my reading of some of that, that that's kind of what happened here. I guess the, the fact that they all happened at once is the thing that scares people. is there anything to be read into that at all, do you think, or do you think it's just circumstance and, coincidence?
Mike MasnickAt this point, I think it's just circumstance and coincidence. I mean, I'm open to the idea that it could be something more, but I haven't seen anything, and it's not like they're not all failing in the same way. It's not-- There's no indication that it's one thing. You know, there have been other cases where, like, Amazon's AWS goes down and half the web goes down, or Cloudflare goes down and half the web goes down. Like those, you can see the coordinated piece that broke. In all of these stories, there was nothing that showed a, coordinated thing that was breaking, and there was no clear indication that it was sort of the same issue with each of these companies. I honestly think it was a coincidence. It may be that, you know, we're at a, point now where, there have been a lot of patches rolling out because of vulnerabilities that were discovered through, Claude Mythos and, I forget what the OpenAI equivalent is. So a lot of people are suddenly, like, going back and checking all their systems for security updates. And we had just a few weeks ago, like, you have these major ones where there's like PHP, which, you know, runs half the internet, had to be upgraded, and that, took down a bunch of sites for a little while. there may be a bunch of little things like that, like people are discovering something saying, "Okay, we need to patch it," and it just timing-wise happened to be like one day after another after another. But I really don't think that they're directly connected in any way, and I haven't seen any indication that that's the case.
Ben WhitelawYeah, that makes sense. I mean, a, a good friend of the podcast, Cat Duffy, who appeared on our live recording, in the summer at Trust Con, she was saying about how she believes, and she's kinda deep in the weeds on this stuff, that it's as a result of the kind of hardening of the defenses, that companies are having to roll out in response to AI. And so there's this kinda like rocky period we're gonna have to go through as companies upgrade their digital infrastructure in order to get to that point. have you had any sense of kind of is this a temporary thing? Are we gonna get through this, or is, or is this the new normal now? Are we gonna kind of experience outages on this scale fairly frequently, because companies are gonna have to adjust to new models, new capabilities that these frontier labs are gonna be rolling out?
Mike MasnickYeah. I mean, one thing I try not to do is predict the future because that is just a guaranteed way to, to be wrong. my sense, if I had to guess, and I, I don't know, please don't, take this as, as like, you know, someone who actually knows what will happen because nobody can predict the future. I would sense that we will see a lot more of this for the near future, but I don't think it is a permanent state of affairs. It wouldn't surprise me if, the next year is sort of checkered with these kinds of examples of big systems failing. but then I would like to believe that once we get through that, we will actually have much, you know, much stronger systems and, deeply hardened systems that handle these things better. Now, it's always tricky because, part of the problem is, like, building something new versus dealing with legacy code. Legacy code is always, complex. And obviously, like, the air traffic control system, as people keep talking about, it's based on these ridiculously old systems, both hardware and software side of it. So upgrading all of that always involves some, very delicate operations, and things can break. but what you hope is that with these new systems, as they're being developed, and if you can take into account getting through that period, that the end result should be stronger, more resilient systems. but that remains to be seen. but my sense is like, yes, we'll probably see more of this for the next, I would say, year or so, but I'm hopeful that once we get out the other side, that we actually do have stronger systems in place.
Ben WhitelawYeah. And, and as well as the stronger systems, I, I wonder if it will bring about a fresh understanding among the public about our digital infrastructure. You know, people who, wouldn't normally think about where, you know, where their emails are powered from and how, and how that works or, you know, the kind of choke points of, a technology company like Apple, are gonna have to s- start thinking about that more, I would say. And, we talk about this on the podcast a lot, that there is no downside, I think, of, more people understanding how the services they use work. And, I think some people will be worried about it, understandably, and it has affected people's work and lives in ways that you know best, or you know more than any this week. But I think as long as people are, kind of engaging the topic, as long as they're doing their bit, as long as they're installing these patches on their devices and, and keeping their security kind of fairly up to date and, telling family and friends about how to do that as well, I think there is, there is a potential upside in the not too distant future.
Mike MasnickYeah. I mean, it'll be interesting. I mean, it's kind of slightly ironic, I guess. The conference I was at last week where I got stuck in Miami was, CloudFest Americas, which was all, like, the infrastructure providers. And so I sat through a whole bunch of discussions just before everything started to break and go down. It w-- That only started when I was trying to get out of there. But, but it was actually really interesting, some of the different sessions where people were sort of trying to understand, like, the way our systems are working. And, and one of the things that came up in one of the discussions that may be relevant now to think about is like one element of the AI boom and, you know, AI touches on this in all different ways, but one element is that people are building more tools and systems themselves. and like I've been doing that. Like I actually, while I was stranded in Miami with, with nothing to do, I actually built a, fun tool that I might talk about at some point. It's, you know, it's in progress, but, I just had this idea and I was like, "Oh, let me build this tool." And I'm hosting it myself. And then part of that is like, well, I have to keep things patched when I'm building tools and then hosting them myself. Suddenly I've become the, the infrastructure security guy as well. And so I'm build- I, I, I was actually building tools to monitor security updates to alert me if I have to patch stuff and things like that. And so I'm suddenly learning a lot more about internet security than in the past, which then leads to like, the reason we have cloud systems and one of the reasons why we have cloud systems in the first place is people decide that it's actually better to outsource that to, you know, organizations that are dedicated to, to making sure that your systems are secure. The reason that a lot of people use Microsoft email, for example, or Google for their email, is so that they don't have to run their own email servers, which is a nightmare. And, you know, keeping those secure is very, very difficult. And so have this sort of push and pull of, who runs the different infrastructure things and, and sometimes it's, you know, there are pros and cons to each of these approaches. And so, it's good when you can outsource it to a large company that will take the security seriously and handle it for you, but then you're putting all your eggs in one basket, and if something goes down, then it can be catastrophic. And so there are all of these different challenges and I think, you know, and this is actually part of the talk that I gave at CloudFest last week, was about, understanding the, pros and cons of that and the value of having control over your own systems versus outsourcing everything. And I think that there are interesting ideas that people are building now, with this idea in mind that, oh, if people are building their own systems, they need them to be secure. And can we build the tools and services to make them secure even if they have more control rather than outsourcing them entirely? And so, we're entering a very different world. we were in a world for the last 10 years or so where almost all of our infrastructure was controlled by a few companies. and that might still remain, but I think there's an opportunity here to shift that so that people have more control over their own services in a way that is still secure. but we'll have to see where that actually goes and whether or not individuals wanna be their own cybersecurity experts and their own sysadmins and everything like that.
Ben WhitelawYeah. And, and there are speech kind of equivalents here and speech parallels, right? Because w- w- we see an increasing number of people, using decentralized social media and, you know, owning the infrastructure or being able to kinda take a kinda modular approach to moderation, to have greater control over that. And that has implications in terms of the maintenance of those systems. It, it creates affordances that actually we haven't, in some cases, haven't really thought through, and, it requires time and energy and capacity and, coordination and organization and, in some cases. And, those are things we, I think are still in the, very early stages of working out. And I th- in many ways, this, these outages are a kind of like signal to me of the work that needs to be done to kinda set ourselves up to, take advantage of, decentralized social media and all that it could offer. it's a kind of, a warning from the future or the
Mike MasnickYeah. Yeah. A-a-absolutely. And I think, you're exactly right is, there are pros and cons to each of these approaches, and we've sort of, to some extent, sleepwalked into the, version that we had. People adopt different things because it's the most convenient solution out there, and we don't necessarily think through the implications fully of that. and there are hopefully better ways that we can do this that give people more control over their own systems. But when you have more control over your own systems, that might also lead to more burden. So, understanding how do we manage that is going to be really important.
Ben WhitelawYeah. Although the people who last Friday didn't have to do any work because their company runs on Microsoft will not care and will, will, will just
Mike Masnickwere, That was, I would say of all the outages, that was the one that made people the happiest.
Ben WhitelawI know. It was the most fun looking on the internet at that time as well. That was the most-- The best memes were about the Microsoft outage, I would say, um, of all of those. Um, interesting. So covered that, I th- I'm h- again hopeful that we won't have to do many more stories about outages in the future, but I think it's an important one to cover. we're gonna go into our kind of eye-catching section now, Mike. We're gonna talk about a few different stories that are worth mentioning. We'll, we'll include the links in the show notes as usual before we move on to our extended Patreon episode. I wanna flag, because it, it is the story of the day in, in many respects across finance and politics and, technology is, Anthropic's IPO. so as of now, on Thursday evening UK time, Anthropic's IPO has, taken place. It has, gone f- very well for the company. it was up as high as kind of 30, 38%, I think, above what it, went on, on the market for. it came down a little bit before the markets closed. But what I wanted to talk about is just the kind of the mention of safety, in Dario Amodei's speech at the Nasdaq opening bell. And we know Anthropic have a reputation for, striving for safety. We've talked a lot about on the podcast, uh, whether that is justified or not, but the, framing of his speech in front of the world's media and in front of the kind of finance institutions is very, very telling, I'd say. He says, The incredible talent of the company that has enabled us to build the most capable AI systems will need to be directed more so than it already has towards double and triple-checking that those AI systems are safe and remain under human control. And to state unequivocally, we will pace the training and deployment of AI systems, taking as long as is necessary to have confidence that they are safe and will remain under human control." Lot of emphasis on, human control, lot of emphasis on, it being a priority. very, very interesting actually, in a world where, as we know, Mike, companies don't necessarily, prioritize safety in the pursuit of growth and, you know, reduce resources, of trust and safety teams and policy teams that are focused on this stuff. So thought that was worth noting. we'll see where that ends up and with OpenAI's IPO also taking place early next year.
Mike MasnickYeah. I mean, you know, obviously this has been a theme from Anthropic from the beginning. It was sort of like the founding reason why they split off from OpenAI, was they thought that they could build a safer version of it. so it's not entirely surprising that that would be kind of a central piece of, the speech that he gave. It's sort of very much tied up with the Anthropic brand. and of course, like we can have debates over what does Anthropic actually mean by safety, and there's all of these issues with like the sort of, arguably cult around AI safety. But like this concept of having a company, you know, a tech company leading with safety as their, first argument is very interesting. you know, especially given the decades that we've gone through with people like Mark Zuckerberg, who seemed to just always view trust and safety as kind of a nuisance and something he had to put up with and, you know, was something that was getting in the way of him maximizing revenue for Wall Street. it is a different sounding thing. I, I still am a little bit skeptical about Anthropic and their approach to all of this. but yeah, I mean, in terms of like a cultural message and the way we think about it and the way we look at it, having the company come out and put safety front and center, is definitely at least noteworthy.
Ben WhitelawYeah, indeed. a story that y- kind of is a change in, a little bit from, from that kind of IPO AI world is, is an open letter that some really interesting organizations have come together to write, that have been published by Snap. Do you wanna take us through why you think that's worthy of, of note?
Mike MasnickYeah. So this just came out, and it's fascinating and, and you can, maybe judge it skeptically because Snap put it out and, they're biased. But it is put together by a bunch of different sort of safety organizations and teenage organizations and teen councils and stuff. And it's, it's really well-written. I suspect part of it may be AI assisted. Uh, but it's a really compelling, from the voice of teenagers, basically directed at policymakers saying that you are leaving us out of this conversation, and you have focused on this idea of banning us from social media, and that is not helpful. And so there are a whole bunch of, like, very quotable sections. I'll just read a couple, small little ones. There's-- one that says, Let's be clear. Teens did not create the current digital landscape. We did not design digital products. We did not write weak laws, delay reforms, ignore evidence, or treat our well-being as a talking point to be used only when politically convenient. And now, after years of adults failing to fix the real problems, a lot of leaders have landed on the same blunt and easy answer: take social media away from us. That is not leadership. That is surrender." it's like, it's pretty strong.
Ben WhitelawYeah. Say, say what you really think.
Mike MasnickAnd, you know, there are a whole bunch of other points in here, and they have, like, actual recommendations for what should be done. And it's basically saying, like, the other part that I really liked was it says, you know, it's an argument that I've made multiple times. So like, having a bunch of kids sort of say the same thing is like, okay, that's, that's validating. It says: "A blanket ban does not teach us how to make good decisions. It does not help us handle risk or build resilience. It does not teach literacy. It does not ensure better product design, and it does not make us safer online. On the contrary, stripping social media away entirely misses a critical developmental window when teens can learn to navigate digital risks under the guidance of parents, guardians, and educators." And it's like, yes, exactly. having them say this, I think is really, really powerful, and I'm, I'm really happy to s- to see this. I think it's r- it's really well done, and well-written and makes the argument that, a lot of us, and certainly, uh, I'm not a part of this group, but the, child safety researchers, who have been screaming about this, the sort of anti, Jonathan Haidt folks saying like the evidence doesn't show. The evidence shows that, you know, if you teach kids how to use these things in an appropriate way, and that's not to let the companies off the hook, like the companies should provide the tools and the, ways to, deal with, things saf-safely. But an outright ban just doesn't seem like a good idea. And so I, I thought this was a really powerful statement, and I was happy to see it. might've been better if it didn't come, wasn't published by Snap itself, uh, because they're certainly, you know, among the companies that are not the most trustworthy on this subject. but it is a very powerful message.
Ben WhitelawThe ones that rely most on those teen users, for the, for their user base. Yeah I mean, it, it links to a story that I also wanna flag, coming out of the UK. we talk about the evidence base being very lacking, and, I think there's been some attempts with the Australia social media ban to kind of assess the impact of, that amongst teens. a number of different academic institutions in the UK have just received a lot of funding to conduct three complementary studies that will track, teens as they go through the social media ban, next year, in the UK. So the UK ban takes place in March. The government has announced that, and those restrictions, are kind of being firmed up through the kind of parliamentary process. But these researchers have, decided that this is a, a perfect opportunity to actually look at the physical and mental effect of being removed from social media, and I think that's gonna provide a lot of, really interesting data and r- research for us, to really know what the impact of these bans are and to know if this letter and these organizations are, are actually right. So, a couple of interlinking updates there. and before we go to our Patreon episode, Mike, you, you wanted to flag a, fun story about, how disinformation takes hold, um, and, and why you shouldn't make jokes on the internet, uh, is, is what I'm taking from it.
Mike MasnickUh, it depends on your definition of fun. Uh, but this was, uh, a blog post written by a, security researcher named Lyra Rebane, who's done a bunch of cybersecurity work and, you know, talks about that they found different vulnerabilities in the past and reported them and, received, various credit for having found them. and so told this story of they had found a vulnerability in Chrome, or Chromium, which is the, the sort of underlying engine that powers Chrome and, and a bunch of other browsers as well, had found a, a vulnerability a long time ago and reported it, and it took a while, but eventually sort of Google claimed they fixed it, released, and then sort of publicly announced that it had been fixed, and it turned out they hadn't actually fixed it, and people started to exploit it, and so they, pulled the bug down. There was, like, this whole mess. And then so they re-fixed it, and then in the process, they accidentally, somebody at Google accidentally added Lyra, the person who had found it, to the bug tracking system as a verifier. So when it gets fixed, someone verifies it, and that should be a third party, not the researcher who found it in the first place. So they, they added her, and that's generally supposed to be, you know, a Google employee, but they gave her those powers. so she started poking around the system and discovered that she had all these new powers, one of which included determining how much bug bounty would be paid. You know, all of these companies have bug bounties. If you find one, you'll get paid a certain amount. It's usually a few thousand dollars. For a bigger one, maybe, you know, $10,000, something like that. So she discovered that she had full access to that, so typed in, $1.33 million. Uh, it's, you know, sort of leet speak, 1.337. It's whatever. It doesn't matter. But, but, like, typed that in, took a screenshot of it, and joked, put a post on Twitter joking that, I got paid $1.3 million, the biggest bug bounty ever." and then immediately afterwards posted a video showing her editing the thing, you know, and saying like, "Or maybe they accidentally gave me edit access," you know, basically joking like it shouldn't be.
Ben WhitelawAnd she wasn't paid that money, was she? Was,
Mike MasnickNo, not at all. Not at all. And, and sort of knew that, there was no way they were actually gonna pay. It was just sort of exploiting a mistake that somebody made at Google and doing it as a, as a joke. The problem then was that, like, lots of people, mainly AI related, totally misread the situation to the point that, like, some people thought that they, you know... You had different AI systems. First of all, community notes, which is a, a cool system, but over the last year, they've added the AI community noters to it, where, like, Grok basically will try and come up with its own community notes. And it misread it and said that the... it did say pr- So it added a community note to her post, and it did say the $1.33 million was false, claiming that she had edited the screenshot, that the screenshot had been faked, which is not what happened. And then you had different AI tools. There are all these, like, AI news aggregators now. So they started saying that she had been paid the $1 million. And then she looked and the, the... If you did a Google search, the AI summary said that she had been paid a million dollars. And then so she wrote this blog post sort of revealing all that, and then found afterwards there's a postscript on it, which is that then all these other AI tools started picking up on the story and even citing this blog post that disproves it as proof that she had been paid the $1.33 million. So, like, you can see how, like, false information spreads, and often it's just sort of like the AI aggregator not understanding sarcasm, you know, and, like, not getting the joke, and that leads to all this sort of false information being spread around. and so, this is sort of an amusing situation. Obviously, like, nobody paid her $1.3 million, and, and, and apparently she was scolded for saying, like, you know, "Look, when you get access to something like that, you really shouldn't change the payment amount. Like, that's pretty bad." Um, but, um, you know, I did think it was an interesting sort of somewhat fun example of, like, where a new form of disinformation may occur, where you have this sort of game of AI telephone effectively, where, like, one sort of joke thing turns into a real thing that turns into a fake thing that then just gets spread really quickly. and, you know, it'll be interesting to see how, how much more of this kind of disinformation as opposed to the, malicious actors or just, you know, whatever. we've seen all these other forms of disinformation that we've talked about for years, but having AI misunderstand something and not understand the context and run with it is sort of a fascinating new vector of disinformation.
Ben WhitelawYeah, and we see so many of these kind of automated news aggregator websites and podcasts and things that, things that don't even have a pair of human eyes look at them ever. there's one example I think of a company who wanted to s- to put out kind of thousands of, of automated podcasts a day. Uh,
Mike Masnickthat's right
Ben Whitelawand, and they've kind of gone back on that model now because obviously it's a terrible idea. But it, it's, it, it's, it is kind of... It's a fun story, but it's also very disconcerting and quite depressing, I'd say. Particularly when, you know, like y- you and I, we're writing, you know, human-created articles and, analysis and trying to figure out the ins and outs of, various speech and, and internet safety stories. Yes, we're using kind of AI, I guess, for, for some aspects of that, but primarily it's human, human done. And like we have editors and researchers and people who look over that, and it's, you know, it's, it feels like you're kind of pushing against the tide a little bit, I'd say.
Mike MasnickYeah. Yeah. and that's what you hope that like, you know... I mean, obviously one of the things that you and I both try and focus on is that we do like careful, nuanced thinking about these things, and that's the kind of thing where AI will often go wrong because it doesn't have the ability to do that, that sort of level of nuance and understanding and pulling in different things. and so, to some extent that is like, sort of reassuring that like we're not gonna be replaced by, by AI anytime soon. But some of that also depends on like the consumers of these bits of information recognizing the value of actually having humans doing the work to determine what is real and what is not.
Ben WhitelawYeah, indeed. And I wonder if Dario in his IPO, uh, thinks there's a place for us, Mike. I don't know. Um, that's probably an, an appropriate time to wrap up our free episode of, today's Ctrl Alt Speech. thanks to those listeners who've tuned in for this. We're gonna turn now to a couple of announcements that took place simultaneously on Monday, one by OpenAI and one by Ofcom, which, are fascinating in their timing and also their significance for what, they both do for detecting child sexual abuse material on the internet. I'm gonna go through kind of a little bit about what they lay out, how they kind of relate together, and then maybe talk through some of the winners and losers of these, announcements as well. so yeah, if you are a Patreon supporter, tune into that now. And if you're not, go to patreon.com/ctrlaltspeech, sign up, and then you can come back and have a listen
AnnouncerThat'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