Ben Whitelaw 0:00

Mike, it's important to remember that large parts of the internet are wonderful. We, we can't lose sight of that. Um, and there's...

Mike Masnick 0:07

a shrinking amount of the internet is wonderful. But yes, yes, sure

Ben Whitelaw 0:11

Yeah. the opening that we do on the podcast is about kind of remembering that and bringing some kind of fun tools to our listeners. what have you kind of come across this week that fits that bill?

Mike Masnick 0:21

So I will admit, this one is a little bit of an oldie but a goodie. I was trying to think through, and this is one that I've known about for a while. I didn't just come across it this week, but it's one that, that always sort of warms my heart and makes me happy. It is the Internet Road Trip. Have you seen this before? Are you familiar with the Internet Road Trip? Oh

Ben Whitelaw 0:39

ne demokrasi?

Mike Masnick 0:40

You are going to love this. it is a website that is built on, you know, the Google, not the maps, the, camera, wherever you're driving. I forget, what is it?

Ben Whitelaw 0:50

The satellite?

Mike Masnick 0:51

Not the satellite. Oh my gosh, I can't believe I'm blanking on the name of it. But, you know, the, the You Are Here. You know, you can picture yourself on the road. You know, they drove around with all the cars and the cameras, and they got all

Ben Whitelaw 1:01

Street View?

Mike Masnick 1:02

Street View! That's the word. Yes. Street View. Google Street View. and so, webpage is, it's part of neil.fun, which just releases all sorts of cool things that are sort of fun to play with, as the domain name suggests. But the Internet Road Trip is just a group project where everybody who logs into it is part of driving a car around wherever you can go, and it has a little steering wheel, and everybody who's on there votes on what happens next. And there's, like, a three-second period where everyone is voting, and you can say go straight, go left, go right. You can honk the horn. You can change the radio station. You can do whatever you want, and there's a little chat going on. Currently, as we speak, there are 126 people driving this car. It appears that we are on a trail, not on a road, in, Lakeside, Canada. it is just really fun. It is an example of a group of people doing something that is totally pointless but totally fun and exciting and, just the, internet of people doing cool, silly stuff. And so, over time, they drive all over the place. You know, every once in a while, I will log in and see where they are. They seem to spend a lot of time in Canada. they seem very, very happy driving around Canada. I, I feel like almost every time I check in, they're somewhere in Canada.

Ben Whitelaw 2:20

That sounds fantastic

Mike Masnick 2:20

is, it is a fun group project.

Ben Whitelaw 2:24

Yeah. A collective of driving enthusiasts doing road trips together. Great

Mike Masnick 2:28

Yeah, or just even like community enthusiasts. There's, there's sort of this community aspect to it where, you know, everybody's just voting every few seconds in terms of what next, and people There is a little chat and people are trying to convince people where to go next. "Everybody vote here," and not everybody agrees, and so you suddenly might take a left turn when you meant to go straight, and so

Ben Whitelaw 2:49

Yeah. It's a good metaphor for life, I think.

Mike Masnick 2:51

Yes, yes, absolutely. Absolutely. What about you? What did you find this week?

Ben Whitelaw 2:56

I came across this funny tool actually called Canvas Browser, and think it's relevant to today's stories because increasingly we're seeing websites and experiences online differ by jurisdiction. I mean, that's always been the case. you know, you see different things if you're in the US or if you're in the UK like me. but this kind of a site allows you to place those different experiences side by side in tiles. So you can say, I wanna see Instagram as if I was in the Netherlands or if I was in the UK or the US," and you can kind of log in and experience it as if you were them. And it's kind of very privacy, receptive, so they don't track what websites you're following or All the tiles that you kind of create are not saved or stored anywhere. So you kind of just see different things, different websites happening in-- as if you're in different countries, which I think is pretty cool and, and relevant to at least one of the stories we're gonna talk about later today. And particularly good, I think, for, big news events as well. If you wanna see how something's being covered in, in a bunch of countries, then it's a fun way of doing that. 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 September the 10th, 2026, and this week we're talking about Australia going after algorithmic feeds, the big AI safety story everyone's talking about, and for our extended Patreon episode, the domain takedown story that's actually flying under the radar. Ctrl Alt Speech is the always linear, never algorithmically sorted podcast co-hosted by Mike Masnick, founder and editor of Techdirt, and me, Ben Whitelaw, founder and editor of Everything in Moderation. Mike, do you think we should be creating a list of our tools for people to follow? I wonder if there's something in

Mike Masnick 4:45

maybe. I mean, people are always sort of curious. There's the, you know, kind of like, you know, what's in your bag or, what is your production function kind of, debate that people have about, you know, how, how people are productive online. So yeah, that's, that's an interesting idea

Ben Whitelaw 5:01

Yeah, possibly. I mean, I don't want to get back to the, the stage where people thought that the cold opens were sponsored by the the company or platform that we were using. But, you know, I never knew about the road trip. I think they're a fascinating selection of, internet geekery. before we dive into today's stories, Mike, I wanna deal with something that a couple of people have raised about the podcast. A very serious matter. Yeah. And it's about the fact we, obviously now doing these, these videos on social media. Um, you can follow Ctrl-Alt-Speech on all major well-moderated platforms, and you can see little, little clips of us talking about the big stories of the week.

Mike Masnick 5:38

Talking heads. Literally talking heads. That's what we

Ben Whitelaw 5:41

That's it. That's it. we-- I wish we were in the same country or city. we have to make do with, being on Riverside every week. A couple of people got in touch, Mike, to, to ask me a very important question, which is, am I wearing the same T-shirt in the last two weeks? And it, I need to put the record straight. It looks like I am. They are two different T-shirts, and this week I've worn a very different T-shirt to make it clear that, I do have a wardrobe that, that goes beyond one

Mike Masnick 6:07

look, there's the Steve Jobs school of wear the same thing every day and you do not have to put any mental thought into it. but, um, yeah, okay. Good, good clarification

Ben Whitelaw 6:20

my, people were questioning my, uh, my washing, schedule, let's say, So good to put that straight. and so we're gonna jump into today's stories. we've got a lot to run through. There is, as we highlighted at the start, a couple of really big stories that have been taking over, I think, a lot of the discussion around safety, particularly AI safety. And so we wanna, introduce a couple of others that, probably are, somewhat flying under the radar, not really being picked up in the same way that we would like. before we do that, we've gotta thank our Patreon supporters, who, do a lot of work to keep this podcast going and, and support us in, in ways that we love. We've got our free supporters who are dabbling in Ctrl Alt Speech, aren't necessarily supporting us yet, but you know, you can become a supporter and see, when the Patreon episodes come out. And then we have our paying tiers. We have our supporter tier that get the extended episodes and earlier episodes for $10 a month. We get our insider and founder tiers who get those longer episodes as well, and the earlier release time, plus the ability to send us ideas for stories to cover, but also episodes that they want us to do as well. So we had a really good suggestion, sent through last week, which we're kind of working on in the background, and you should see more of those in the coming months if you are a Patreon supporter. So go and check that out, patreon.com/ctrlaltspeech. Mike, we, we're a global, podcast. We've gotta, we've gotta start with the global story of the week, which I think is about what's going on in Australia. how kind of familiar or how closely you've been following what's going on in, in Oz over the last few months?

Mike Masnick 7:47

Yeah, I mean, definitely very closely. Australia, I used to joke Australia was the place where good internet regulations went to die. it just seemed like they were quite literally the upside down land in terms of their thinking on internet regulation as opposed to elsewhere, and I think that has maybe only continued. But that also means that there's sort of, an ongoing work in progress case study that we can look at and observe their efforts in, in progress and see, well, you know, look, maybe I'm wrong and maybe they will one day magically be successful. We've certainly talked about their leading the world in, banning social media in some form for those under 16, which I would argue has been a pretty dismal failure. and they're sort of, instead of maybe going back and looking at why that was a bad idea, they're just sort of charging forward and saying, "And the next thing."

Ben Whitelaw 8:41

Yeah

Mike Masnick 8:42

and so what they've now announced is, specifically this, duty of care concept, which has certainly been talked about elsewhere. it's part of the debate over KOSA in the US, it's part of the debate in the UK and in the EU and lots of other places, this idea of implementing a duty of care. And so Australia is moving forward with that, including, Well, we'll get into the issues with duty of care in a second, but the sort of big headline that is getting lots of attention is that they want to include this, it's not quite algorithmic choice feature, but this algorithmic rejection feature, which is that they want to force all social media to give users the option to turn off the algorithm and only get a chronological feed. and that's sort of like the, big headline, and in fact, the announcement from the Australian government was literally titled "My Feed, My Way," which sounds very grand and, and very exciting. but I think there are a lot of questions around this, and so let me dive into, a couple of the issues, and then we can, go deeper in, in whichever direction you'd like. so the broad duty of care is one of these things that I think And this comes up a lot in a lot of the discussions that we have. It sounds good, right? Like, who doesn't want a duty of care? And who doesn't want these companies to take these issues seriously? And who doesn't want them to, be thinking about their users and protecting their users and keeping their users safe? Obviously, all of that is good. In practice, what a duty of care is, is really a censorship regime because it is a sort of vague standard which is saying, like, you have to put in place things to protect your users while they're using your platform such that nothing bad happens to them. But it is impossible, literally impossible to predict what bad things may happen to users of your platform, at some point, in some way connected or not connected to their usage of the platform. Part of the issue is that when you have billions of people using these tools, sometimes bad things happen to those people, and sometimes bad things happens while or in some way connected to their usage of the platform And a duty of care really structurally is designed to punish a company afterwards for not having the magical predictive foresight to figure out every bad thing that might have happened, so that as soon as anything bad happens, the government can come back to the company and say, "You failed the duty of care because you did not predict this thing," which now in retrospect seems wholly obvious that you should have figured out and you should have prevented, even though if you're trying to prevent everything, you're going to miss other things and not really understand what is happening. And on top of that, if you are trying to predict any and all harms to any and all individuals, what you end up doing is just taking down all sorts of content because the risk is too high, the liability is too high. And this is, you know, and I've made this comparison before, but this is literally how the Great Firewall of China worked in the early years. When China wanted to block certain content, the way they did it was rather than saying, "You have to block this kind of content," they went to the ISPs and they said, "We will hold you liable if anything bad happens on your service. We will not define what is bad. We will just say we will hold you liable." So what do the companies do? They just say, "We're just gonna take down anything that might possibly be considered bad." And obviously what happens in that case is you over-block because you're just trying to protect anything, possibly bad from, happening in, in any way, shape, or form. And so the duty of care concept tends to lead to mass over-censorship and platforms just trying to be as clean as possible. Now, some people might argue that that's the world they wanna live in, sort of a Disney-fied world in which, you know, anything negative or dangerous is, is not seen. But lots of other people say, well, that also means that important discussions, controversial discussions, views that challenge authority, all of these things get wiped out along with that.

Ben Whitelaw 12:48

And, and just to be clear, Mike, the, the duty of care concept you're talking about here is kind of fact that it doesn't apply to speech in the same way that it does to, other products and services, right? Because the concept of duty of care exists in other industries. It ex-exists in the real world. You know, it's a kind of-- it makes sense, from a kind of consumer protection point of view, which is, which is where a lot of platforms are having to defend their products, increasingly, because of the harms that they have supposedly caused. But you're, you're making the point that, that it doesn't map so neatly onto speech.

Mike Masnick 13:19

Exactly. and that's a really, really important and good clarification, so thank you for, for, for holding me to that. Yes. in all sorts of physical world, products, you expect a duty of care. you know, if you're making food or whatever, you don't wanna poison people. If you're building a car, you don't want something that's gonna, be faulty and crash and kill people. But when you're talking about speech, and internet, social media is inherently a speech product, every bit about it, the code itself is speech, the users on it are engaging in speech, it becomes much more difficult because what is harm when you're talking about speech? This is not to say there is no harm, but it's a very different form of harm, and the same kinds of speech can impact people in different ways can have different impact. And if you're saying, like, any harm from speech is bad, that leads to really bad situations because then somebody criticizes somebody else and suddenly you're like, "Oh, that harmed me." That creates really, really big problems in a variety of ways too. So using a duty of care concept around speech generally just leads to censorship and/or disproportionate enforcement in terms of who and when do you enforce that rule. you know, it basically depends on who's in power and who they wanna punish, and that becomes really, really problematic in all sorts of ways that hopefully listeners of this podcast could predict.

Ben Whitelaw 14:38

Sure. And so let's talk about some of the specifics of, of these draft amendments then that Australia are looking to push through under the guise of the ch- kind of duty of care. And so kind of Online Safety Act, as many will know, came out in 2021. it's, you know, been tried to be enforced by the Safety Commission and people-- There are lots of kind of very similar obligations on platforms as we've seen in other jurisdictions around transparency, around, re-reacting to, to harms on platforms. and so these, new aspects are kind of layered on top. And, the ones we- you talked about, the algorithmic feeds one is the headline, but there are other aspects as well, aren't there? there's, kind of obligations for platforms to avoid, certain features being used by under 18s that, and you're gonna love this, have negative behavioral impacts is the phrasing. And so there is this kind of assumption, particularly in, in regards to features, like chatbots and, the risks that come with AI as we, as we increasingly know, that people know what those negative behavioral impacts are or that we have this shared view of what they are. And I guess your point you're making is that we don't necessarily know that, and it's it's, different depending where you are and, and what government

Mike Masnick 15:49

and, right, and not only that, but like at the individual level, like different people are impacted in different ways, and the same features may be helpful to some people and harmful to others and, and neutral to many others. There's so much of an assumption built into all of this that there are certain features that are inherently harmful to a vast majority of people, and therefore they need to be away with, or that we can introduce new features that will magically be helpful to more people when the research has not supported that. And so, like, that leads right into the, the feed issue, right? This idea that the algorithmic feed is the problem is underlying a lot of this, and that's why they wanna take it away and force companies to offer a chronological feed. And I'm not against... Like one, I'm entirely in favor of algorithmic choice. I want people to have more control over their algorithm. I've been pushing for that for years. I mean, that was a central part of my protocols, not platforms paper. It's a central part of what Blue Sky does with their algorithmic feed and choices. Ben is reaching for The bell. Ring the bell. There we go. I'm on the board of Blue Sky. Take anything I say about Blue Sky with a grain of salt or many grains of salt. I believe in algorithmic choice. I believe in empowering the user. But this is entirely based on the idea that the algorithm is the problem as opposed to the lack of individual empowerment. So they claim that this is individual empowerment. They talk it up the same. This is not the government taking control. where it is, 'cause the only thing they're trying to offer here is a chronological feed as the alternative to the algorithmic feed. And, research has sh- shown that, like, that doesn't seem to help. In fact, you know, if you are just using the chronological feed, you tend to see more disinformation, more harmful content, because the reality is, and I- a lot of people don't seem to recognize this, the algorithm is a safety tool in many, many cases and often protects you from bad or harmful content. And yet what they're saying here is, do away with it as if that will help. Other studies have shown, like if you talk to people within these companies who have worked on this stuff and who have offered people chronological feed, the reality is like the vast majority of people don't use it and don't like it. They like the algorithmic feed. And there are always people who get really mad when I point that out, but it is the reality. People find that the algorithm is actually helpful

Ben Whitelaw 18:20

And, and in fairness, the Australia's Communications Minister, Anika Wells, in her announcement about this initiative and about the duty of care did, say that more so than I expected. You know, she said people do like algorithmic experiences. It's not gonna be the case that everyone, should move to a kind of chronological linear feed. But what we're wanting platforms to do is to offer you know, an alternative experience to users on a regular basis. And we don't exactly know what that looks like. Maybe that's a pop-up. Maybe that's some sort of not- you know, notification you get in a very rarely seen part of your settings that you never actually venture into. we don't know what that looks like. and I think y- you're right to say that the research is fairly, you know, wonky on this, and we would include a really good piece by, the Pro Social Design Network who, who review some, some academic research. And, and to your point, Mike, they find that actually the down-ranked experiences, the algorithms where they down-rank bad content

Mike Masnick 19:18

Yeah

Ben Whitelaw 19:19

actually are looked to be the kind of most strongest experience for users, which is exactly your point you're making, right? That people don't realize that it's not just content being surfaced that is engaging or addictive or, good for people, but there is also algorithmic kind of weightings given to bad content, and that content is kind of sent down the pecking order, so to speak, and, and won't be visible to users. so, so my question on the back of that is like, why do you think algorithms have, have only been seen to be promoting content rather than down-ranking content? what do you think has kind of led us to, in the common conscience, to think that?

Mike Masnick 19:53

Yeah, I mean, I think it's a few things. I think that one of it is, like, it's much easier to wrap your head around the things that you do see as opposed to the things that you don't see. It's just like there is this sort of cognitive situation where you don't consider the things that you never see. and so there's just that sort of inherent instinctual you're, you're thinking about that. The second is just sort of like the narrative that many people had about, social media and the algorithm over the past 10 to 15 years, where there was a period of time when, social media algorithms were first becoming popular and the companies were experimenting with them, that they did experiment with more rage bait style, algorithmic things because they believed that, like, that was engagement. Like, they sort of over-optimized and over-indexed on engagement, and that included engagement based on anger. the reality is that most platforms very quickly realized that that was not an effective tool and actually drove people away and caused longer term problems. It may have gotten people more engaged in the short term, but it burnt people out. And so, YouTube very, very quickly realized that. Meta, Facebook at the time, they realized it fairly quickly. They had done some research. This was some of the, documents that were released by Frances Haugen, that I think were also misrepresented, talked about how Meta experimented with effectively a rage bait algorithm. but then you look at the details, and what they realized was like, "Oh, this doesn't work. Like, this is really, really bad, and we have to change it." And they changed it really, really quickly. was like a few months long that, that algorithm was sort of, focused on that. but, like, that sticks, you know, the, the narrative around it, really sticks. And in fact, like, I had criticized, this is many years ago now, I had criticized, there was a Washington Post article that talked about that aspect of the rage bait algorithm within Facebook that, I can't remember how far into the article, mentioned that it was no longer Like, it only lasted for a few months, but the headline was like, oh, Facebook specialized in, encouraging rage bait. and so, narratively, that's the kind of thing that sticks, and then people think back, like, when they first got on social media, it was fun and engaging, and back then there was no algorithm. Therefore, it must be that it's the algorithm that is the problem rather than a bunch of other factors in the world.

Ben Whitelaw 22:15

Yeah. I mean, I, before kind of we, came on air, I started to use-- I went on Instagram, which I never really do, apart from to look at clips of Ctrl-Alt-Speech, of course,

Mike Masnick 22:24

everybody should do. Yes,

Ben Whitelaw 22:25

which everyone should do. and I, and I switched to the, following option, which you can do. I know you're, you're not a big Instagram user, but y- there is, you know, it's a very hidden feature, but you can hit the, the banner at the top and you have this following option, you have this favorites option. You go to the, following option, and it's a completely different experience, not least because currently, and this isn't good product design, Instagram don't offer any of their other features when you're using the chronological feed,

Mike Masnick 22:54

Ah.

Ben Whitelaw 22:55

which I think is not something that's very well known. you know, so they're kind of actively degrading the experience as a result of my choice to use a, chronological feed, a non-algorithmic feed, which I think is, is interesting. And I think what Australia is trying to get around is that you can't, degrade an experience if you want to use a different, a different feed. There is precedent here, which again, I didn't know before we started kind of researching this story, that in, in the Netherlands, you can actually pick a for you feed, on Instagram as a result of a 2025 court case that was brought as a result of the DSA. That means that, there is no essential default, feed experience for Instagram users in the Netherlands. You, pick either the algorithmic one, or you pick the kind of chronological one, or you pick your favorites feed, which is people you've selected. and I found out as well that that experience maintains every single time you open the app,

Mike Masnick 23:50

Right. also it doesn't degrade the other features. So whereas what you found is you go in and you wanna just do the chronological feed, then that removes like the ability to DM, which is a weird setup. But that doesn't happen in the Netherlands because of this court case. And s- so that's an interesting experiment, and I'd love to see the data on sort of Instagram usage in the Netherlands since that has happened. there's some indication in the article that, you had pointed out that like people still don't realize that, that this is the way it is set up in the Netherlands, and they maybe got a pop-up once and they've forgotten about it. so there's all sorts of things about like notification and how do you get people to understand this stuff. And again, like I do think that user empowerment is a really important tool, that not just, allowing the companies to make all the decisions for you, is really important. But, part of that is also like how do you make users aware of the power that they have, and how do you get them to use it wisely? And that's a whole other discussion that, you know, we haven't really approached yet because, most people don't even have these options yet. But if we're going to start forcing these options on people, we should have much clearer, whether it's media literacy, digital literacy, whatever it is, there has to be, much more of an effort and, much more, interest paid to, you know, how do we get people to use these things wisely?

Ben Whitelaw 25:08

Yeah, exactly. And that, came up in, in the recent Meta settlement and some of the kind of disclosures, around that was that safety features exist. there are parental controls that exist. There's a lot of thoughtful work being done into some ways that people can augment their experience, but people don't know about it and they're not told about it. And so y- know, it's almost there for a regulator and not essentially for a user, which I think is probably what Australia's trying to get around here a little bit

Mike Masnick 25:34

Yeah. And I, you know Again, like, if you look at the details of the, Australian proposal now, there's just a whole bunch of stuff that, like, is is the kind of stuff that feels good and sounds good if you don't actually think about it for more than 10 seconds. So, you know, I was looking at, the requirements for protecting young people from harmful content, which again, sounds wonderful, right? You don't want kids to experience harmful content. But they have this thing that says, like, Content that promotes, encourages, or accelerates eating disorders." And I've written about this a bunch, and there have been a bunch of studies about this. That is one of those things that sounds wonderful, right? We don't want content that encourages eating disorders. The reality is that is so difficult because there were studies done that Instagram, this was now, 2016, 2017 timeframe, I think, if I remember correctly. There were these studies done where Instagram was getting hammered because there was sort of pro-eating disorder content that was found on Instagram. And so Instagram cracked down hard on that. And did that improve eating disorders among young people? No. What happened was that people were looking for it. There was a demand problem, not a supply problem. So people continued to look for that kind of content, and they found it in, like, various forums that were way less well-moderated than Instagram ever was. And what they had found was that prior to this, on Instagram, where there was content like that, you also had people going in and offering tools and support and services and s- suggesting like, "Oh, here are helpful resources." And some of that was Instagram provided, like, you know, PSA-type content. but the really powerful thing, and the thing that was most impactful, was you had people on Instagram who had had eating disorders and had recovered from them, who were then showing up in the comments and sort of helping guide people who were going through eating disorders to resources and help and recovery tools. And that all went away when Instagram was sort of told, like, "You have to crack down on this." And so the, the wider impact of saying like, "Oh, you have to block this content," wasn't actually helpful for those with eating disorders. and so, these are really, really complicated topics. And also, you know, one of the things that they found was, like, as Instagram tried to crack down, people just started using different language, and then it just becomes this cat and mouse game. So, like, there was a whole discussion about, like, people were posting photos of chewing gum, and then there was an argument that that was, like, eating disorder content. And at some point you're like, "Well, how are you determining what is actually eating disorder content versus what is literally just a photograph of, of a pack of gum?" and so it's nice to say these things, but the reality is, is that it's way more- tricky and way more nuanced than, politicians like to make it out to be

Ben Whitelaw 28:25

Yeah. And certainly very hard to capture in a few words of a new draft of a already, already long and complex bill. We've got a bit more to say on Not Really, but we're kind of rapidly running through today's episode. We've got a couple of small stories we wanted to, talk through. these are things that are kind of eye-catching stories to us, which we're gonna flag to listeners to go and read in our show notes, and then we'll move to our Patreon extended section. Mike, one of the things that you almost can't not have seen this week if you've been on the internet is this huge story, I'd say, although listeners may say otherwise. I always kind of wonder if we're, in the weeds on this stuff more than others. But there's this big story about an AI researcher, at Anthropic who's quit and written a series of posts on X, essentially saying that we're at risk of extinction as a species within a decade

Mike Masnick 29:13

optimistic.

Ben Whitelaw 29:15

And it has blown up in a big way. as of this afternoon, I think it had 150 million views. it's done the rounds in every single news agenda, and this, young, unassuming 27-year-old Brit called Jacob Coxen is now the center of discussion about AI safety. there's almost too much to unpack here, and we, we kind of presume that most people know a bit about it. But you wanted to kind of highlight why this is, this is maybe not quite as big a story as it should be

Mike Masnick 29:43

I mean, it is an interesting story, right? And, so there are a h- a number of different things, right? People... There have been sort of the AI doomer class that has been warning about this kind of thing for years, and then you have the opposite side of it that says, like, "Uh, this is just an exaggeration." Or you have I mean, it's funny because the, the criticism comes from both directions, right? You have some people who are saying this is really just marketing for the AI. "Ooh, we've made a tool so powerful it k-kill all humanity. You should pay us 20 bucks a month to use the tool to help you with your work." You know? Um, and then you have the people who are saying Whatever, this is complete garbage. It's, it's useless." you have all sorts of, different viewpoints about, how real this is or how serious this is. But there is something interesting happening here in, in my mind, which is, you know, we've seen, obviously we, we spoke about the hugging face incident, on our live episode at Trust Con a f- a couple months ago. and there's been a lot more release since then about these AI agents, creating message boards basically to plot which, you know, is very science fiction and, and very scary. And you're saying like, "Oh, if they're doing all of that without the AI labs being aware of it, what else might they do?" if they get on task and they decide that getting on task requires them to do crazy things that may be harmful, will they do it? And this has always been the, the issue of so-called alignment within the AI space. and I think that you have this, combination of these factors that are all sort of hitting at once, and then this guy who, is somewhat young and not that experienced and maybe not fully an expert on all of these things, but is basically saying that there's a, a strong belief within Anthropic that, total disaster is likely. That's going to get headlines and people are going to pay attention. And, you know, part of what I think is interesting about this is that for years, the backlash in the social media space that we've, discussed and talked about was that the people building these tools didn't take safety seriously at all. They were building tools to free the world and give everybody a voice and blah, blah, blah. It was gonna be, all this wonderful bring freedom and democracy to the world. and then what people complained about was like, "Oh, but you didn't think through the safety implications of this," and now we're seeing that, that there are some safety implications of it. In the AI space, we're seeing somewhat of the opposite, where you have all these people who are building the tools saying, "Oh, by the way, this could kill off humanity

Ben Whitelaw 32:10

Right

Mike Masnick 32:11

and we should be thoughtful about that." And you have some of the same people who are complaining about social media folks not taking this stuff seriously sort of mocking the people who are taking it too seriously in the AI space. And the reality in both cases may be somewhere in the middle, where it's like you need a balance of sort of trying to build the best tools possible that are really useful and are really helpful, while also keeping in mind the potential safety downsides of them, and then trying to figure out a way through. But all of these are complicated by the fact that, like, nobody, can predict the future properly and it's very difficult to predict where the real problems are versus the ones that are purely science fiction. And so, I think this is just sort of, this is going to keep boiling up in some form or another, and this is just sort of the latest example of that.

Ben Whitelaw 32:57

Yeah. it's a fascinating story and partly because of the, the reaction to it. I, I saw a BBC presenter in an interview, with somebody who basically exclaimed, "Oh my God," it live in an interview because it kind of hit her that there was this potential kind of existential threat,

Mike Masnick 33:14

yeah. I mean, that's an interview with Geoffrey Hinton, who is, you know, often considered the godfather of AI and, you know, as a professor, taught many of the, the leaders in the AI field and was, worked as a leader in the AI field. And for a long time he's been sort of sounding the alarm about the risks and dangers and, and he's sort of, often lumped into the doomer camp. But he's in the, that camp that is a knowledgeable doomer. and he said in that interview that, made her say, "Wow," or, "Oh my God," he's like, "You know, it's kind of impossible to predict the actual likelihood of something catastrophic happening." But, you know, he accepts the idea that it might be 10%.

Ben Whitelaw 33:52

Yeah. It's about 9.9% more than I would like at this

Mike Masnick 33:56

Yeah, it's, it's, it's a little unnerving, sure

Ben Whitelaw 33:59

yeah. but yeah, definitely include some, some, some useful reading material in the show notes. There's also a-another kind of eye-catching story that you wanted to quickly f-flag, Mike, which leads us into our Patreon episode, about an individual who we've talked about in other episodes in the past who's, been on the receiving end of some, let's say kind of interesting US government treatment

Mike Masnick 34:18

Yeah. S- so this is Imran Ahmed, who's the founder and CEO of CCDH, which is the Center for Countering Digital Hate. and I've talked about in the past, like, you know, they do a lot of advocacy work and research work around trust and safety issues. Again, I assume a lot of people listening to this podcast will be aware of them. I am generally less than impressed with their output. I tend to think that their research is not very good, their methodologies are problematic and not particularly realistic. and so there's a part of me that feels that they've contributed a lot to the moral panic around technology and internet speech that I think is misleading and unfair. So I'm not a fan of the organization in general. however, the attacks on them are ridiculous. So Elon Musk sued them a while ago, because they had done one of the early reports about, ads for big brands showing up next to Nazi content. and they've done other research that makes X and Elon Musk's control of X look bad, and so he sued them. For the most part, they have won on First Amendment grounds. But at one point, the Trump administration, and Marco Rubio in particular, famously, I think we also talked about this on the podcast, decided they were going to ban certain, European censors from coming to America because they were harming free speech.

Ben Whitelaw 35:46

including your friend Terry Breton

Mike Masnick 35:48

Breton is, is, was one of the, five people. But, Imran was one of them, and he was unique among the five in that he is actually in America, even though he is not a US citizen, but he does have, legal residence here and is married to an American and has an American child, and has been living here legally. but as part of this, Rubio, basically determined that he is, uh, you know, against the interests of the US, and by designating him as such, opens him up to the possibility of being deported, which is clearly an attack on his speech. As much as I disagree with his speech, the things that he is saying and the criticism that he has made of social media and their content moderation practices is his free speech, and it should be protected under the First Amendment. And so he sued to try and block any attempt to enforce, and to enable his own deportation. And unfortunately this week, a court rejected that lawsuit and basically said, that Imran may have fully legitimate arguments and his speech may be fully First Amendment protected, but under the law, the court has no jurisdiction over this at this time. This is just like, it gets kind of wonky and into the specific, legalistic weeds. But basically, Congress passed a law that limits how much courts can, get involved at the investigation stage. Like, if the attorney general or, you know, a cabinet official opens an investigation or a procedure about someone, that's too early. And so there's an argument that it's, it's just too early. But the reality is that, like, when it gets to the point where they're actually trying to deport him, then suddenly it's a lot tougher to do things because things are, are in action. So this is clearly a campaign against him for his speech. It's, I don't know if it's ironic is the right term, but it is like an attack on his free speech that Marco Rubio is claiming is to protect American free speech, which is just garbage. and so, Ahmed says that, he's fighting this and will appeal, and you know, it, that it's not gonna silence him, which is great. Like, speak up, speak your mind, you should be able to do that. But this is, part of the current US government's attack on people thinking about trust and safety, and attacking them for their speech and claiming that their speech is censorship. I think it's noteworthy that, he may face being deported, for his speech,

Ben Whitelaw 38:15

Yeah.

Mike Masnick 38:15

is just very, very troubling.

Ben Whitelaw 38:17

and one of the effects that we, we've heard from trust and safety researchers and practitioners about these five individuals' cases is, is that it's kind of stopped them from doing their work because, because of the ramifications. and we already know that, and that's, that's the kind of, this has been a long, longstanding kind of legal battle, and this isn't gonna help with that at all. so, so it, it doesn't bode very well for people who are, who are doing this work in academic institutions or, or within platforms, frankly. You know, if you think you're going to be deported for doing your job, then that's a very

Mike Masnick 38:49

and your job being protecting people. And, like, the idea that this is just about censorship is just garbage. but, you know, we spent all these years where people built it up as if, all trust and safety was just censorship. and this is the, the end result of that. So it's, it's horrifying

Ben Whitelaw 39:05

Yeah, it is. it does lead, neatly onto our extended Patreon episode. We're gonna wrap up our free episode now. We're gonna go into a deeper kind of content moderation story, Mike, which is a kind of, frankly, quite frightening story about how, I'm afraid it's the US government again, designated an organization in Italy as a terror organization and essentially kind of shut them down within a week. Talk to us about kind of like why this story applies to Ctrl-Alt-Speech listeners, why

Mike Masnick 39:30

so I-- this is a really important story, and it's gotten very little press coverage. and basically, this is an Italian organization that set up tools for hosting and messaging and email and, messaging that was used by lots of activist groups. and because the US government doesn't like the fact that certain activist groups were using this to communicate and that they didn't have access to it, they decided that they could designate it as a terrorist organization. There's no real due process or no nothing, and that has caused a bunch of American companies that they relied on to cut them off. And that is, huge and incredibly damaging and, is something worth considering about, like, how does the global internet work when the US can effectively turn off an entire service based on the say-so of just Marco Rubio?

Ben Whitelaw 40:21

Yeah, exactly. Everyone, everyone who has a domain, who works for a company, who is at the same level of risk as this Italian organization should tune into this. Uh, we're gonna dive straight in. Thanks, Mike.

Announcer 40:32

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