Age-Gating the Internet + Cloudflare Takes On A.I. Scrapers + HatGPT

“Frankly, if they’re going to behave like hackers, then we’re going to behave like trolls right back to them,” said Matthew Prince, chief executive of Cloudflare.

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:00Speaker 00

  Support for this podcast comes from Dell Technologies. It's time to put AI to work. Is your business ready? The challenge isn't getting started with AI. It's scaling beyond pilots while maintaining control of your data, governance, infrastructure, and costs. Dell AI Factory provides the integrated foundation you need to scale AI securely and efficiently with one trusted partner, Dell, AI on your terms. From idea to impact. Learn more at dell.com/yourwaytoAI. That's dell.com/yourwaytoAI.

• 0:35Kevin Roose?

  People on social media, Gen Z people, have started referring to robots as clankers. Have you heard this?

• 0:42Casey Newton?

  I believe I saw something about this.

• 0:45Kevin Roose?

  Yes, so people are saying like, oh, like, I hate when I call customer support and a clanker picks up. And, you know, this is sort of their new derogatory slang.

• 0:54Casey Newton?

  Well, and it pairs nicely with another new piece of slang that I wonder if you've heard, because people are talking about people who use tools like ChatGPT a lot, and they're calling them sloppers.

• 1:03Kevin Roose?

  Really?

• 1:03Casey Newton?

  Have you heard this?

• 1:04Kevin Roose?

  No.

• 1:04Casey Newton?

  Yeah, so it's like, if you're, I don't know, you're out on the internet and you're publishing something, and it seems like it's obviously just AI, somebody might call you a slopper.

• 1:12Kevin Roose?

  Hmm, well.

• 1:14Casey Newton?

  Clankers and sloppers.

• 1:15Kevin Roose?

  Clankers and sloppers. I just think this is, it makes me a little uncomfortable, even though clankers is not, you know, actually, I think it's sort of a tongue in cheek thing. I just don't think we should be calling the robots names. I don't believe in slurring against anyone.

• 1:29Casey Newton?

  You believe in an appeasement strategy with the robots. Just give them what they want.

• 1:34Kevin Roose?

  They are keeping score.

• 1:35Casey Newton?

  I believe that.

• 1:41Kevin Roose?

  I'm Kevin Roos. I'm a tech columnist at The New York Times.

• 1:44Casey Newton?

  I'm Casey Noon from Platformer.

• 1:45Kevin Roose?

  And this is Hard Fork.

• 1:46Casey Newton?

  This week, why age gates are suddenly popping up all over the internet, and how some could create more problems than they solve. Then, CloudFlare CEO Matthew Prince returns to the show to discuss his company's new plan to help websites fight back against AI scrapers. And finally, we're passing the hat for some hat GPT.

• 2:17Kevin Roose?

  Well, Casey, we are here yet again to remind our great listeners that they can have their very own Hard Fork hat if they so choose.

• 2:25Casey Newton?

  It's time to get the hat. If you don't have the hat, the moment is now.

• 2:30Kevin Roose?

  Yes. For a limited time, you can get a Hard Fork hat with the purchase of a new annual New York Times audio subscription. This is a special edition hat. This color is not going to be available anywhere ever again. And it'll be going on eBay for hundreds, possibly thousands of dollars.

• 2:48Casey Newton?

  Millions of dollars.

• 2:49Kevin Roose?

  Millions of dollars. And you'll be like, why didn't I get a Hard Fork hat? Back when it was available to me as part of a New York Times audio subscription.

• 2:57Casey Newton?

  If you missed out on the Bitcoin boom, get this hat.

• 3:00Kevin Roose?

  Yeah, they're calling it Hatcoin. And it's available now at nytimes.com/hardforkhat.

• 3:12Casey Newton?

  Kevin, how old are you?

• 3:14Kevin Roose?

  It's none of your business. How old are you?

• 3:16Casey Newton?

  That's none of your business. But guess what? If we lived in the United Kingdom, it would be the government's business, Kevin. Because as many people found out over the last week, to use a lot of different websites, you now have to prove your age.

• 3:28Kevin Roose?

  Yes, this is the age gating issue, which I've heard a lot of people talking about. I know you wrote about it in your newsletter this week. I'm excited to talk to you about it. And before we get into what is happening and some of the consequences and the reactions to this law, I wonder if you could just kind of sell me on the stakes. Why talk about how the UK's Online Safety Act is mandating age verification for websites? Why does it matter to me, an American?

• 3:57Casey Newton?

  Sure. So I would say a couple of things. Number one, this truly is one of the most far-reaching attempts we have seen by a Western democracy to regulate speech online. We've talked on this show in the past about various ways that people want to protect children online. And requiring folks to verify their ages is one of the ways that has been discussed. But it's extremely rare to see it roll out across an entire The way it has in the UK. So that's thing one. Thing two is this stuff is coming to the United States. In June, the Supreme Court upheld a law in Texas that requires residents of Texas to do something very similar if they want to access adult content online. And so it's really not just a UK thing, Kevin.

• 4:41Casey Newton?

  We are starting to see a gradual erosion of people's freedom of expression online.

• 4:47Kevin Roose?

  Yeah, I think that's well taken. And I might just even take a further step back and say like, I think the internet for the last, call it 40 years, has been a kind of informational free for all, right? You were going to experience the same internet whether you were 13 years old or 25 years old or 50 years old. But what we are talking about now and what is happening in the UK is you have to establish your age in order to sort of access a number of different kinds of services, including social media. And one outcome here would just be that the internet actually kind of fragments in this way where if you are 12 or 13 years old,

• 5:30Kevin Roose?

  you are just going to have a very different experience of the internet than someone who is 25 or 50.

• 5:36Casey Newton?

  Yeah, and I think in addition to all of that, Kevin, there is just the fact that the way that websites are collecting this information is putting people at risk in ways that we can talk about.

• 5:47Kevin Roose?

  Okay, so let's just start with what is happening in the UK. What's the backdrop here?

• 5:52Casey Newton?

  Yeah, so in 2023, the United Kingdom passes something called the Online Safety Act. And it includes provisions that require online services to try to spare minors from seeing what they would call harmful content online. So porn is a big part of that, but it also includes stuff about like if you have a pro-suicide website or a pro-eating disorder website, you are now required in the United Kingdom to first do a sort of like risk assessment of your website of, hey, could children access this? And would they be exposed to this certain list of harms? And if so, you then have to implement what they call like high-quality age assurance, which means that No, you can't get away with just putting a box on your website that says, yes, I promise I'm 18.

• 6:36Casey Newton?

  And so all of this goes into effect last Friday, the 25th, and things start to go a little bit haywire.

• 6:42Kevin Roose?

  What happens?

• 6:43Casey Newton?

  Well, first of all, I think a lot of people in the UK just didn't realize that this was about to happen. And so, you know, they're sitting down for their evening visit to Pornhub and all of a sudden, they find that they are being asked to upload their driver's license to prove that they're actually 18. There are a number of different ways that people are allowed to prove their identity. You could also, for example, show your credit card because you can't have a credit card unless you're at least 18. Or you can use your phone or laptop's camera to take a picture of you. And they'll, they'll do some sort of, you know, AI in the background to figure out if they think that you're actually 18. But for a lot of adults who didn't see this coming, this produces some real anxiety.

• 7:25Casey Newton?

  Because all of a sudden, an experience that had previously been basically completely private, right? Just visiting a website is now being linked to your personally identifying information. And who knows what's happening to it once you actually submit it.

• 7:39Kevin Roose?

  Right. Now, is this just affecting porn sites? Or are there other kinds of websites where people are being asked to prove that they're of age?

• 7:45Casey Newton?

  It is affecting many more kinds of sites. So X is being affected. Many different subreddits are being affected. So like a stop smoking subreddit, a cider subreddit. Various things that do not immediately seem like adult content and could actually be quite beneficial to minors. All of a sudden, you need to prove who you are if you want to access them. Also, Wikipedia has said that they may have to limit access to the site in the UK because of privacy concerns that are created by this law that, you know, would require them to collect a lot of information that they don't actually want to collect.

• 8:21Kevin Roose?

  Now, is there porn on Wikipedia?

• 8:23Casey Newton?

  Uh, I haven't been able to find any. Have you?

• 8:26Casey Newton?

  Well, I will say there are some Greek vases with some incredibly curvaceous men and women. And, um, you know, depending on what kind of night you're having.

• 8:35Kevin Roose?

  Right. So how are people in the UK reacting to this?

• 8:38Casey Newton?

  This is not a keep calm and carry on situation, Captain. Okay, more than 400,000 people have signed a petition saying that they want these changes to be reversed, and I'll be curious to see where that goes. In the meantime, though, people are just finding new ways to get around this. And I think to the extent that this law actually sticks around, I think this will be the reason why, is that people have just found all sorts of relatively easy workarounds. One thing you can do is use what they call a virtual private network, or VPN. All that does is just lie to your internet service provider and say, hey, I'm not in the UK, I'm in the United States where I can still watch most of what I want to watch. And so we're seeing a lot of that. But we're also seeing people get much more creative.

• 9:22Casey Newton?

  Now, are you familiar with the video game Death Stranding?

• 9:24Kevin Roose?

  No, I'm not.

• 9:25Casey Newton?

  So unfortunately, this is so funny, if you have, if you know what Death Stranding is, it's this kind of very strange and beautiful video game made by this Japanese auteur. And people are using photo mode in the game to take pictures of the protagonist in order to fool the UK's age verification technology. And the reason this works is because in the photo mode of this game, you can tell the character, smile or frown. And it turns out that a lot of the age gates in the UK are doing the same sort of thing. Where it's like, if you're using the camera on your laptop, you're like, okay, now smile, okay, now frown. People are now just doing this with the game Death Stranding. So if Death Stranding winds up being the best-selling video game in the UK this year, we'll know why.

• 10:06Kevin Roose?

  That's amazing. So, okay, has there been any response from the UK government or the safety authorities about people's reactions to this? Are they saying, we'll keep improving the systems or something like that?

• 10:18Casey Newton?

  There was a response from Ofcom, the UK's media regulator. They have said that this act is not, quote, "a silver bullet," but, quote, "until now, kids could easily stumble across porn and other online content that's harmful to them without even looking for it. Age checks will help prevent that."

• 10:34Kevin Roose?

  All right, so that is what is happening in the UK. Where are we in the US with these age-gating laws? I've read some headlines about people in some states who are having to go through age verification to get to an adult website. But where are these laws? And what is the regulatory picture here?

• 10:53Casey Newton?

  So, age verification laws have been passed in 24 states. And last month, the Supreme Court upheld a law in Texas that requires websites where more than one-third of the content is sexual material to use one of these age verification methods. And Justice Clarence Thomas spoke for the majority when he said, "Unlike a store clerk, a website operator cannot look at its visitors and estimate their ages without a requirement to submit proof of age. Even clearly underage minors would be able to access sexual content undetected." So, I suspect we're going to start to see these laws in more and more places.

• 11:26Kevin Roose?

  So, all right, let me try to steelman the case for age verification here, because I think I am undecided about what I think about it. But I think you have decided that these laws are a bad idea. So, I want to try to make the opposite case. We age gate things all of the time to prevent minors from getting access to it or being exposed to it. So, if I walk into a bar and the bouncer asks to see my ID, that is allowed. No one has contested whether that is a violation of the Constitution. If I walk into, this probably doesn't exist anymore, but when I was a kid, there were these things called nudie magazines. And if you wanted to go into a newsstand and get them, they were sort of like behind the counter. They would have these like special barriers on them.

• 12:11Kevin Roose?

  And you would have to prove that you were of age to be allowed to buy that. So, why is what's going on with age verification on adult websites on the internet any different?

• 12:21Casey Newton?

  So, I share your concern here. I think that we should come up with ways to prevent minors from accessing this kind of adult material. I just object to the way in which they're doing it, which requires that people share a lot of personal information for the rest of their lives.

• 12:36Kevin Roose?

  So, you're not like opposed to age-gating in, like, as a concept? You're just opposed to the way that they're doing it in the UK?

• 12:42Casey Newton?

  Yes, and this year, Apple... Suggested a way to do age verification that I think is a lot more elegant, Kevin. It's not available quite yet. They said this week to Bloomberg that it's coming soon. But basically, they're going to offer what they call an age assurance API. And if you are a parent and you're setting up your child's device, you can just tell the device, hey, my kid is X years old, right? Or here's my kid's birthday. Apple can then essentially anonymize that and pass it through to a developer. And so if you are a Facebook or an Instagram or a TikTok, you just get a little token that says this person is 13. And so, you know, you may want to show them different kinds of content or you might want to restrict certain features. And essentially, put the onus on the the parents and the device makers to do this.

• 13:27Casey Newton?

  That way, if you're just a normal adult using the internet, you don't have to worry about uploading your driver's license to visit a website. So I think that that's a very elegant solution.

• 13:35Kevin Roose?

  Yeah, I prefer the on-device age verification rather than making, you know, every website operator go and like, do their own version of this and store all the driver's license photos and whatever else. So I did not know that Apple was building that. So you're saying that's going to be released soon?

• 13:51Casey Newton?

  Yes, they're they're going to release that soon. Now, interestingly, Meta is trying to ensure that this does not become the way that this is handled, because that still puts an onus on Meta to do a lot of the age checking. They want Apple and Google, which also has a big app store, to have the legal liability in cases where a minor does access material that they're not supposed to. And so Meta has been leading a charge lobbying for a lot of bills around the country that would put the onus on Apple and Google to do all the verification instead of them having to play a role in it. And they've been having some success.

• 14:26Kevin Roose?

  Hmm. Now, this confuses me because, you know, if I know one thing about Meta and other social media sites is that they are very good at collecting information on users and using machine learning to sort of detect who is more interested in what and who is part of what consumer segment. I assume that these platforms already know or have very good guesses about how old all of their users are. So what is the issue with just? Having them do the detection. I also saw that YouTube this week is looking at things like your browsing history and your consumption patterns and the kinds of videos you're searching for to make their best estimate of whether you are underage

• 15:09Kevin Roose?

  or not. So why shouldn't the platforms have a responsibility or a role here too?

• 15:14Casey Newton?

  I think the platforms do and should have a responsibility. What we have seen is a lot of reporting over the past couple of years that at Meta in particular, they were like writing reports about how many like under 13 users they had of Instagram. Jeff Horowitz in the Wall Street Journal did a lot of this reporting. It's you know, both very disturbing, but like darkly comic to read. And like, yes, they absolutely knew that they had all of these younger users. And so they've spent the last year trying to release a lot of features that essentially make it more difficult for, you know, under 13s to use the platform. So all the platforms are kind of belatedly coming around on this. You know, you mentioned the YouTube thing. Here's something really interesting about YouTube. This week, the Australian government said that they were going to ban YouTube for kids under 16

• 15:55Casey Newton?

  because they consider it social media, which was a major reversal from, from what they were saying before. I sort of think this is the sort of thing that if it was actually carried out, could cause the government of Australia to be toppled.

• 16:06Kevin Roose?

  A bunch of like angry Minecraft pre-teens are gonna storm the, whatever the, the White House of Australia is.

• 16:12Casey Newton?

  Yeah, but, and we'll never know, but it just goes to show you, actually, if you, if you know what the seat of government is in Australia, please email [email protected]. We were unable to find this information online due to age checks. Anyways, look, I've been a little bit, you know, glib about this, maybe in the spirit of trying to record an entertaining segment about tech policy, but the, the truth is that this stuff is very complicated. And I think everyone has a role to play, which is something that is very satisfying to say, but does not actually solve the problem. Because ultimately, if you want to solve the problem, somebody has to be responsible, somebody has to try to implement a solution. Inevitably, when you implement a solution, people are gonna be caught up in it. They're gonna be, you know, falsely tagged as underage when they're overage or vice versa. So it is really messy.

• 16:56Casey Newton?

  What I'm trying to say is there are more privacy preserving ways of understanding a person's age online. And I would like To see us focus on those rather than this sort of blunt hammer approach that they're taking in the UK, which in practice is mostly just going to annoy a lot of adults and potentially put their public information out there in a way that could be breached, which is my way, Kevin, of introducing the conversation about tea.

• 17:22Kevin Roose?

  Oh, yes. Let's spill the tea about tea.

• 17:23Casey Newton?

  Let's spill the tea about tea. So this is a story. And by the way, are there any red flags about you on tea? I haven't looked yet.

• 17:30Kevin Roose?

  I have not been able to look, nor will I be looking because of what happened over the past week. So tea is an app that had a viral moment over the past week, in part because it reached briefly number one on the iOS App Store, ahead of ChatGPT and Instagram and all these other apps. And to my understanding, although I've never been on it, tea is an app that allows women to sort of anonymously divulge experiences they've had with men.

• 18:01Casey Newton?

  Yeah.

• 18:01Kevin Roose?

  So basically dish the dirt about the guy you dated who was sort of a jerk to you or behaved in a way that you didn't like. And what got people's attention was that you could only register for this app if you were a woman. So they would do some kind of verification process when you signed up. I think you were asked to like to scan your driver's license and also submit like a selfie so that they could sort of use AI to try to detect whether you are a woman or not and keep out all the men. And this drove a lot of people on the internet insane. And so my understanding is they essentially hacked

• 18:39Kevin Roose?

  tea. They they found that tea had not secured some of these uploads that users had made. And they leaked a bunch of people's verification information and photos that they had submitted to tea and kind of a revenge of the men.

• 18:57Casey Newton?

  Yeah. And this gets to my exact concern with some of these age gating methods is it is just left up to the service provider to decide how you're going to verify people's ages. And while most countries do have some laws that regulate data, in practice, we just see breaches. All of the time, it feels like every week, you know, you see headlines about one app or another is being breached. And in this case, you have material that is just ripe for someone to do stuff that is really abusive. And indeed, a bunch of 4chan users got a hold of all of these T-selves. They started creating like, you know, Hot or Not style websites, just doing all kinds of gross stuff with them.

• 19:36Casey Newton?

  And that again, I think is just a really predictable outcome of building these kinds of age gating technologies.

• 19:43Kevin Roose?

  Yeah, I think that's a really good point and an example of why we should not be throwing this to the website operators and platforms. Because inevitably, some of them are going to use really secure, really well-designed age verification services. But some of them are just going to do this in a very sloppy way and like inevitably leak people's personal information.

• 20:05Casey Newton?

  Yeah, and now like every online service just has to like have this database laying around. You know, it just like increases the risk across the entire internet for all of us.

• 20:13Kevin Roose?

  Yeah, so I think it makes a lot of sense to do it in a more centralized way through Apple and Google and their app stores. But I have also read that Apple has been lobbying against some of these state bills that would force them to do exactly that. So you just told me that Apple is building an age verification feature. But I've also read that Tim Cook, at one point, personally lobbied the governor of Texas not to do this kind of age verification thing. So what is going on here?

• 20:42Casey Newton?

  So I think it's mostly just that they want to avoid the legal liability here. You know, I read one story that said like, look, like, Apple is a mall, Facebook is a liquor store in the mall. Like, you need to hold the liquor store accountable for who they are selling liquor to, okay? Like, this is like the Apple metaphor. I think Apple does clearly want to play a role in the solution here. That's why they've sort of crafted this whole API. But they kind of want it to be on the honor system, right? Where it's like, hey, we're going to do this, we can all agree this is a good thing, now leave us alone. And what the lawmakers are saying are like, yeah, that's like a good thing, but also we want to hold you legally liable if we find out that like a bunch of, you know, eight-year-olds are, I don't know, like, you know, watching eating disorder content. TikTok.

• 21:25Kevin Roose?

  Yeah, I think that makes a lot of sense. And I'll also say like, this is an issue, this age gating thing, where I find that my values are in tension. Because I agree with you that the First Amendment is important, that adults should be able to like, access all manner of information on the internet, even if it is lewd or obscene or inappropriate for children. I don't want a kind of UK style system where like, you have to upload your driver's license to some, you know, database that may or may not be secure, just to go like, look at a subreddit. At the same time, I totally understand that parents have legitimate concerns here.

• 22:10Kevin Roose?

  I talked to so many parents who are just at their wits' end about how to safely introduce things like social media, or smartphones, or the internet to their children, without just sort of opening the floodgates. And parental controls are the thing that everyone talks about, but any parent can tell you, these are not perfect. You practically have to become like a part-time IT person to just to be able to like, understand and control what your underage kids are doing on the internet. So I totally understand where the anger comes from at these companies that have made it very clear that they don't want to be doing any of this. They don't want this to be their responsibility.

• 22:52Kevin Roose?

  But I think that we are going to move to an internet with age verification in the United States and around the world. I just think that is going to happen. There's enough public pressure at this point that it is sort of inevitable in my mind. And so I would like to see it done in a way that preserves the privacy of those users, rather than having sort of T-style leaks every six months.

• 23:11Casey Newton?

  I would like that too. You know, here's sort of the last thing I would say about this, Kevin. You know, this may not be the free expression that you, the listener, cares about the most. You may be happy to upload your ID if you want to look at adult content on the internet, or even Wikipedia. But what I am going to say is, I view this very much as part and parcel with just a widespread Clawback of speech rights in the United States over the past six months. Look at the attacks on free speech that we've seen against journalists, against academia, against broadcast media, right? And now you are coming along and saying, by the way, now you need to show government ID to look at a website. Like, these things are all of a piece. And I think it's important that we think about them as being of a piece

• 23:56Casey Newton?

  because it may be that, you know, six months, a year from now, we look back and we think, well, we have really lost a lot of ground with free expression. And the time to talk about that online is while you still can.

• 24:07Kevin Roose?

  I see. Well, thank you for explaining this to me. It was not immediately clear to me why I needed to care about a tech regulation in the United Kingdom, but I think I get it now.

• 24:18Casey Newton?

  Yeah, I feel like just doing this segment has aged me about 10 years.

• 24:23Kevin Roose?

  Yeah. Also, I'm looking at a list of British expressions that mean really angry. And I just have to read some of these too. Like a bear with a sore head.

• 24:34Casey Newton?

  Okay.

• 24:35Kevin Roose?

  Throwing a wobbly.

• 24:37Casey Newton?

  Throwing a wobbly?

• 24:38Kevin Roose?

  Yeah. Anyway, I'm throwing a wobbly over this regulation. And I think a lot of other Brits are in a right tizzy.

• 24:48Kevin Roose?

  When we come back, Matthew Prince from CloudFlare will explain how he's trying to keep the internet safe from AI crawlers.

• 24:55Casey Newton?

  Not to mention the clankers and the sloppers.

• 25:19Speaker 00

  Support for this podcast comes from Dell Technologies. It's time to put AI to work. Is your business ready? The challenge isn't getting started with AI. It's scaling beyond pilots while maintaining control of your data, governance, infrastructure, and costs. Dell AI Factory provides the integrated foundation you need to scale AI securely and efficiently with one trusted partner. Dell, AI on your terms. From idea to impact. Learn more at dell.com/yourwaytoAI. That's dell.com/yourwaytoAI.

• 25:53Speaker 03

  Building and scaling AI today can feel pretty chaotic, mostly because teams are juggling too many fragmented tools at once. That's where Outsystems comes in. Outsystems gives you one unified agentic systems platform to manage your entire portfolio of agents, apps, and workflows. You get to innovate at the speed of AI, but with the security, governance, and control your enterprise actually needs. And the results are already there. A major manufacturer used Outsystems to build a multi-agent system that analyzes blueprints, and a global bank cut their development time by 60%. Own your agentic future today at outsystems.com/nyt.

• 26:30Speaker 04

  AI spend is growing fast, but most businesses still can't see exactly where that money is going. Ramp gives businesses visibility and control across AI providers with real-time guardrails by team and model, helping ensure spend goes to the highest ROI areas. Stop guessing where your AI spend is going. Get started with Ramp. And if you're new to Ramp, you can get $250 off at ramp.com/hardfork. Go to ramp.com/hardfork. Ramp.com/hardfork.

• 27:03Kevin Roose?

  Well, Casey, we've got a very exciting guest with us today. Matthew Prince, the CEO of CloudFlare, is here to talk about something new that they've been working on.

• 27:11Casey Newton?

  Yeah, and I am so excited about this, Kevin, because it actually is kind of in the dead center of my own interests as somebody who has a website on the internet.

• 27:18Kevin Roose?

  And Casey, because we are talking about AI in this segment, we should talk about our disclosures.

• 27:23Casey Newton?

  That's right, Kevin. Well, my boyfriend works for the notorious AI scraping company Anthropic.

• 27:28Kevin Roose?

  And my employer, The New York Times, is suing OpenAI and Microsoft over alleged copyright violations related to the training of large language models. So CloudFlare, for those who may remember our last podcast with Matthew, is sort of the plumbers and bouncers of the internet, right? They make a lot of these security services like DDoS protection that many, many websites rely on to keep their services safe. They also do a host of other security and data-related things. And Matthew is a sort of diehard supporter of the internet. And so a couple of years ago, when all these AI bots started scraping Internet to feed data into their models. He and CloudFlare got really nervous about this.

• 28:08Kevin Roose?

  And so they have been doing a lot of interesting things to try to preserve what they see as the the heart of the Internet. And recently, they made a lot of waves by announcing that CloudFlare would start to introduce default blocking of these AI data scrapers, basically preventing AI companies from automatically being able to exploit and scrape the data from the websites that they're visiting. This is a step that he said would help to protect content creators like you online.

• 28:37Casey Newton?

  Yeah, and so I'm excited to talk to him about that. And also, Kevin, just about some of the research that CloudFlare has been doing into the state of the web as a whole. You know, you and I have both been worried about what the rise of AI means for the Internet in general. CloudFlare has taken a close look at that and has some interesting things to say.

• 28:53Kevin Roose?

  Yeah, it's one of the sort of more underappreciated companies on the Internet. Their decisions about even seemingly small things can have enormous ripple effects throughout the industry. So I thought it was a really good time to bring on Matthew and ask him some questions about what they're doing and why.

• 29:10Casey Newton?

  Well, let's bring him in.

• 29:18Kevin Roose?

  Matthew Prince, welcome back to Hard Fork.

• 29:20Speaker 05

  Thanks for having me.

• 29:21Kevin Roose?

  So we're going to talk about AI and scraping and all the steps that you're taking to sort of protect the Internet. But before we describe the solution that you're proposing, which you announced on something called Content Independence Day, I want to ask you to describe the problem. You had some pretty shocking numbers in your post on this about how much Internet traffic patterns have already changed as a result of AI. So talk a little bit about those and why you felt the need to do something about it.

• 29:52Speaker 05

  Sure, so, you know, at CloudFlare, we sit in front of north of 20% of the web. So we see a lot of what's happening online. And we understand, I think, a lot of how the web works and the the business model that exists behind it. And 25 years ago, basically, Google struck a deal, an implicit deal with publishers, which was, you let us copy all of the content that you're creating, and in exchange, we'll send you traffic. And the whole web, the economy of the entire web, was really built on that search-based interface. However, over the last 10 years, Google has changed what that interface looks like.

• 30:36Speaker 05

  It started sort of subtly, starting about eight years ago when they introduced the answer box. And the answer box is if you type in something like, when did Cloudflare launch? Instead of you having to go to 10 blue links, it just says September 27th, 2010. And that was a pretty radical change because back in the day, Larry and Sergey used to brag about how their job was to get people off google.com as quickly as possible, right? They wanted, they even measured the time up in the corner to show you just how fast you basically were, you know, leaving their site. And now all of a sudden, Google was keeping you on the site. More recently, they've introduced something that's AI overview. So if you do a search, most of the time now, overview shows up, which is an AI summary of what's going on.

• 31:19Speaker 05

  And that has made a big difference in how much traffic goes to content creators from search. How much of a difference? Like, put some numbers on that. Yeah, so if you, if you take 10 years ago as the sort of the standard and you compare it with today, it's almost 10 times harder to get traffic today for a piece of content than it was 10 years ago. And it's gotten significantly worse just since the introduction of, of AI answer box. That's just with Google and that's the good news. If you look at the AI companies, OpenAI is 750 times harder, Anthropic is 30,000 times harder.

• 31:59Speaker 05

  What's basically happening is as the interface of the web is shifting from search to AI, people are consuming derivatives, not originals. Basically, they're not following the footnotes. And we've actually seen it get even worse as people are actually trusting the AI systems more because the AI systems are getting better. And what scares us about that is that there are really three reasons people create content. Maybe, maybe not you, Kevin, but one is ego. Vanity is mine. Right? But that's, that's a big thing. Like, again, there's a whole bunch of content that's just created, you know, for just the ego and the fame of, of, of doing it. Yeah. But the business model is you either sell subscriptions. Or you sell ads or both of those things.

• 32:43Speaker 05

  And so the value has to be either make people famous or make them rich through ads or subscriptions. That is what has built the web over the last 25 years. The problem is that publishers are struggling already at 10 times harder. I worry that as more and more of the interface of the web looks like OpenAI or Anthropic, they're dead at 750 times and they're buried in the ground and forgotten about it 30,000 times. And so if that's the case, if there's not an incentive for people to create content, I think people will stop creating content and that really is an existential threat to the web.

• 33:16Kevin Roose?

  Okay, that's a pretty good and concise description of the problem. So what is the solution here? What are you all doing?

• 33:21Speaker 05

  Well, so I don't pretend to know exactly what the solution is, but I know kind of some of the aspects of how we have to get there. So the answer has to be that AI companies have to pay for content. The deal is different than it used to be with search. Search, they copied your content and in exchange they sent you traffic that you can monetize. If now they're copying your content, they're not sending you anything, then why would you give them their content in the beginning? So AI companies have to pay for content. And I think what's encouraging is we're actually seeing some AI companies that are doing that. Amazon just announced a deal with the New York Times, that OpenAI has done a number of content deals that are out there.

• 34:03Speaker 05

  And but the problem with a lot of those deals that we realized was they, you know, Sam at OpenAI can't be a sucker. He can't pay for content and then have all of his competitors get it for free. Or another way of thinking is you can't have a market unless you have some level of scarcity. And so what we thought the first step was and what we announced on July 1st in conjunction with the who's who of the world's publishers, was we're going to by default block the ability for AI crawlers to get content unless they are compensating the content creators for getting that content. And we think that that's super important. Now, exactly how the compensation works, again, I think there's a lot of different models. I analogize it to, and at risk of hubris,

• 34:48Speaker 05

  when Apple announces, you know, the inter- Of iTunes 99 cents a song, right? That wasn't what the final business model was. The final business model that we've kind of come to is more like Spotify, where it's 10 bucks a month and you get kind of all you can eat from the Spotify playlist. I think we're going to take some iterations to figure out where we, we might start at some price, you know, fixed price that's there. And we might evolve to something which is more like a Spotify model over time. But the first step has to be actually saying, if you're an AI company, you can't get content for free.

• 35:22Kevin Roose?

  Now, if I'm a website operator and I want to prevent AI bots from scraping my site, I can already do that through robots.txt. I can just put a little file in the metadata and I can say, hey, these three crawlers, you're allowed to scrape, but these other six, you're not allowed to scrape. So how is what you are building with CloudFlare different than that?

• 35:44Speaker 05

  Well, the first thing is with robots.txt, the analogy would be like it's the speed limit sign, where it says, okay, you can drive 55 miles an hour, right? There's no law of physics that says you have to draw 55 miles an hour. And I think a lot of us maybe sort of look at the speed limit sign and go, ah, yeah, it's probably 60 is okay, right? It's the same thing. It is a recommendation. It's not an actual enforcement that's there. That's problem number one. Problem number two is it's incredibly blunt in terms of what it does. So you basically have to apply it across a significant portion of your site. So you can't say, okay, let these things out, but not these things. The last problem is a lot of the search companies, because again, we see a huge amount of the internet.

• 36:29Speaker 05

  We can track how this works. A lot of the AI companies, I should say, if they hit a robots.txt file, what they then do is they find some other way to go out and scour the internet to find your same content. So they'll actually do a search against Bing and try and pull the cache content. Or they'll go look at the internet archive. Or they'll actually do things that are incredibly sneaky, like pinging ad server networks to get a description of the page that comes back to them and trying to find all kinds of things around it. We've tracked some of the sort of worst performing AI companies. And there's a huge range like OpenAI. They're actually the good guys here. They're doing things right. They're trying to do things the right way. But by far the best behavior.

• 37:10Speaker 05

  There are others that that most closely resemble like North Korean hackers where they're literally using residential proxies to spoof who they are to try to get around around the various blocks. And so I think that in addition to a road sign for the badly behaving bots, we actually need something where we say, we're going to take away your car because you keep, you know, going 400 miles an hour down the road.

• 37:33Casey Newton?

  Tell us a bit how technically the solution that you've built works. How are you able to build those kind of fine grain controls?

• 37:41Speaker 05

  Yeah. So I mean, so CloudFlare, we're essentially a giant network and a ton of the web sits behind us. And so it has to flow through our pipes. And our primary business for most of our history has been cybersecurity. So every day we go to war with the North Korean hackers, the Russian hackers, the Iranian hackers, the, you know, Chinese hackers who are trying to get into our customer's systems one way or another. And we're really good at identifying, no matter what they pretend to be, we're really good at identifying who they are, what they're doing and stopping them. Literally by just not letting their traffic get to our customers.

• 38:19Speaker 05

  What we realized, though, was that actually gave us the perfect position to be able to help anyone who is a content creator also set what were enforced rules of the road. Where we can say if if this bot is behaving in a bad way, then we're going to block it. So for instance, we, you know, as we have now studied this and we have very good evidence on what even some of the major AI companies are doing that's really sleazy. We're going to put them in timeout. We're actually going to stop their ability to access a huge portion of the web, even even if they, you know, pretend that they're doing it. And we've got the record to do that. And our security teams have actually investigated and figured that out. On the other hand, when, you know, somebody has, if OpenAI has done a deal.

• 39:00Speaker 05

  We also want to make sure that they get that content as efficiently as possible and structured in a way which is which is possible. So the content creator can say, I want to allow OpenAI to come to my my page. I think what will develop over time is then what is sort of a standard rack rate where you as a content creator, even if you're small, can say, okay, I'm happy to let this content be scraped. But here's the price for it. And that's something that, you know, how that market will develop, I think, is going to be what, you know, kind of the next steps of of this project.

• 39:30Kevin Roose?

  Now, you announced this new approach to AI crawlers a few weeks ago and flipped this switch that blocks all the crawlers by default. What has the reaction been from publishers, from AI companies, from people who are inside Cloudflare looking at the data?

• 39:45Speaker 05

  Yeah, so I mean, from publishers, not surprisingly, I'm on a lot more people's Christmas lists. Like, we're, we, this, because publishers were really struggling with how to deal with this. They saw the problem and they didn't have a good technical solution to be able to lock it down. They're, you know, they're not cybersecurity companies, so they have a harder time being able to track this. And so the glee that we've seen from publishers as they've turned the system on, seen what was going on, and then had the ability to push a button and say, you know, disallow, disallow, disallow, disallow, has been really, really palpable. And I think that's why this resonated across such a wide swath of the publisher base.

• 40:26Speaker 05

  What surprised me has been the reaction of the AI companies, which I kind of thought that they were just going to throw up all over this and hate it, but it hasn't been that. For the most part, with a couple of exceptions, they've said, listen, we get it. Ultimately, content is the fuel that runs our engine and we need to pay for it, but the key is, it needs to be a level playing field. And so I think that what I am encouraged by is in all the conversations that I've been a part of, the AI companies said, if you can make it a level playing field, if you can make it something that's fair, then we're willing to pay for content.

• 41:06Speaker 05

  And that, I think, is, that's a dramatic step that says that we're headed in the right direction. Now, making it fair is going to be a trick in folks like Google, who can just believe that they've got kind of a God-given right to be able to copy everything off the web. Like, that's going to take some persuading to get there. But I think as the industry lines up and says, this is the right thing to do, that we'll be able to even get Google to, hopefully voluntarily, but if not, there's certainly enough investigations on them going on around the world that, one way or another, I think that they will be persuaded or compelled to get behind this effort.

• 41:46Casey Newton?

  I mean, basically, what you're describing is a revenue share deal, right?

• 41:49Speaker 05

  That's right.

• 41:49Casey Newton?

  And it seems only logical that a company like Google, which is going to be making a lot more revenue by crawling everyone's content because they're such a source of queries, should be paying more than the Upstart AI company that just shows up on day one and wants to start crawling the web. So I would hope that they would be open to that.

• 42:07Speaker 05

  I think I am encouraged and I do believe that Google really does believe in the ecosystem and they get it. But it's a little bit of the frog boiling in water. Where 10 years ago, or 25 years ago, when Google, when Google started, it was a good deal. Like, get in the nice pot of water, you're happy, you're a frog, right? But they've slowly turned the heat up and that's made what used to be a good deal into a much, much worse deal. And so the deal needs to get renegotiated and that's a big piece of it. The other thing that I think is important is you need to have access to content. And that's the thing today that's cheap. But my prediction is that over time, the real differentiator between the different AI companies is who has access to the most interesting content.

• 42:51Speaker 05

  And we've seen this play out with things like Netflix, where you can see that by getting original content, it actually drives subscribers. And so I think that at the end of the day, it's exactly right that fundamentally what content producers should be arguing for is we should be getting a share of whatever the revenue that you're generating from users is because we're the fuel that runs the engine that is powering your business.

• 43:14Casey Newton?

  I feel like you can already see this today. And Kevin, I'm sure you've seen this. Matthew, I'd be curious if you have too. But if you ever like use one of these chatbots to run a deep research report because you're trying to like really, you know, bone up on some, you know, set of historical facts or, you know, recent current events that you want to write about, you get these results and like half of them are from like bleepbloop.com, you know, like newswire.xyz, like publishers that no one has ever heard of and that you're not entirely sure are on the level. And so often when I run those searches, I think, God, I wish that the sources I trust actually did just have deals with these chatbots where that I could like log in with my like Bloomberg credentials and then just have you read Bloomberg too as part of the report that you're making.

• 43:57Speaker 05

  Yeah. And I think, I mean, all of those things will come, but two things. One, bleep blop.

• 44:04Speaker 05

  bloop.com actually sometimes is going to have really interesting original content. And I think part of the key here is we don't want to create a situation where the only people who get the content deals are the major, major publishers. I mean, the New York Times can do a big deal with OpenAI. At the same time, you want to make sure that the small AI companies, the new startups are able to also get access to this. So, so a vibrant market here is lots of sellers and lots of buyers. And you want to, you want to be able to do that.

• 44:32Kevin Roose?

  Right. Now, you mentioned that publishers are very happy about the steps that CloudFlare has taken to block these AI crawlers, that AI companies were not as excited understandably, but, but maybe they would get on board too. I did talk to one AI executive about this, who basically accused you and CloudFlare of setting up a new toll booth between the AI companies and the content companies, because you are not only providing this technology, but you are sort of inserting yourselves as the merchant of record in these transactions. And this person asked me to ask you, what percentage of each payment CloudFlare plans to take? So what percentage does CloudFlare plan to take from these transactions?

• 45:12Speaker 05

  So first of all, like if you're the New York Times and you do a deal with OpenAI, that's your deal. We don't get any of that. If we facilitate that, so if we're, we're the ones who go out and negotiate it, then yeah, we'll take some percentage of it. And I have no idea what that will be, but I think it'll be something reminiscent of like what spot, what does Spotify take of a subscriber's revenue versus what do they pay out? And usually that's sort of in the 20 to 30% range of, of what that is. I think the only way that we should get something is if we're actually generating value from this. And, and, and if not, then you should do the deals yourself between the AI companies and the publishers. And in that case, sure, we'll provide you the interface to, to be able to stop it.

• 45:56Speaker 05

  But, but that isn't something that we would take any, any percentage of. What's also important is we provide this at no cost to even our free users, because we think that this is fundamentally important to the sort of health of the long-term internet. Like it shouldn't be something that just the big companies can, can get access to. And so no matter who you are, if you're signing up for CloudFlare, you get these tools, you get the analytics, you get the understanding, you get the ability to block it. And again, I think that we will be less involved in the transactions for folks like the New York Times, but for bitboploop.com or whatever, that's probably, it's probably a porn site that we're pointing people to.

• 46:34Speaker 05

  And you can sort of envision what it might be. But if that's the case, then yeah, I think that that's a place where if we're doing work, then we should get compensated for that work.

• 46:43Casey Newton?

  I mean, you know, so I have a website on the internet, platformer.news, that's where my newsletter is, and I'm, I'm super interested in this because I frankly just do not have the time or energy to go out and try to strike deals with AI companies. I also have no idea what the value of Platformer is in that particular marketplace. So if somebody wants to go like make a market and then I can just show up and you tell me like, hey, you know, we can make you 80 bucks this year, or like whatever it is, like I'm interested and you want to take 20% of that, sure.

• 47:11Speaker 05

  That's, yeah, and I think that that's, that, that feels fair. And again, hopefully we're not the only ones doing this. Like from my perspective personally, like my wife and I own a small newspaper in, in our hometown, we see how hard and how important it is to have local news. And, and there's got to be a business model to it, like it, that reporters need to eat, you know, it costs money to print papers, like you have to have a business model that's there. You know, I think personally, we've built a $60 billion company on the back of the internet. Like I feel an enormous responsibility to give back to the internet and actually protect it. Our mission is to help build a better internet. If you talk to anyone at CloudFlare, that's why they work for us. And so like, with all due respect to the AI executive, they've built an entire company

• 47:54Speaker 05

  by stealing content creators' content and not compensating them for that. And if they keep doing that, people will stop creating content, which not only kills the internet, but kills them in the process.

• 48:05Kevin Roose?

  Yeah, I think there are a lot of people who would agree with you. One of them is not President Trump. He said just last week, in connection with some new AI policy rollouts that he was doing, he basically took the side of the AI companies and said that they shouldn't have to pay for copyrighted material. He said, quote, "You can't be expected to have a successful AI program when every single article, book, or anything else that you've read or studied you're supposed to pay for. It's not doable." So I guess, what is the incentive for an AI company to agree to something like a pay-per-crawl system if the current administration is signaling that they're not going to face any penalty for just breaking copyright the old way?

• 48:45Speaker 05

  Yeah, and I think, like many things that sometimes come out of the Trump administration, nuance is a little bit lost here. We've actually talked with a number of people in the Trump administration, and I think what they are concerned about is much more a government-instituted, kind of, "You must pay X" in the way that we've seen come out of Australia and the way that we've seen come out of Canada. I think that's what they're trying to signal that they're not in favor of. They are very much in favor of, as far as we can tell, private market solutions where that is created. And so, again, I think the incentive for the AI company is, you need content in order to have access to build your tools, and we've just stopped you from getting it.

• 49:28Speaker 05

  And so regardless of what the law says, we're going to be able to technically stop the AI companies from being able to get the content unless they're compensating the content creators for it.

• 49:37Kevin Roose?

  I want to ask you, Matthew, about another release that you all did earlier this year related to this issue of AI crawling, which was something called the AI Labyrinth. This was a system that was designed to, basically, instead of blocking AI crawlers, it would just sort of redirect them to an endless series of AI-generated links and pages, basically trapping them in this labyrinth that they could not escape from. So my first question is, do you worry that the robots will remember that you did this to them and take their revenge?

• 50:06Speaker 05

  It is not a risk factor that we have currently included in our S1, but I'll talk to our legal team about it. We now have the data that there are some AI companies, big AI companies, well-funded, that are just behaving horribly. And frankly, if they're going to behave like hackers, then we're going to behave like trolls right back to them. And we can feed enough garbage into their system that they will create garbage content. And so again, when we come back to what's the incentive, like you don't want to piss us off. Because, because again, we believe in the future of the internet. We believe in supporting journalists. We believe in supporting content creators. We're on the side, the right side of history here.

• 50:49Speaker 05

  And so the right thing to do is say, yeah, you're spending $10 billion on GPUs. You're spending billions of dollars on employees. You should be dedicating at least something to paying for content.

• 51:00Casey Newton?

  Did the labyrinth work? Like, what was the result of that?

• 51:03Speaker 05

  It's, I mean, for real, again, we try not to put like OpenAI, good, good actor, doing great, doing all the things right. So we, we don't throw them into.

• 51:11Casey Newton?

  This message was not endorsed by the New York Times legal team.

• 51:15Speaker 05

  That's, yes. But, but for bad actors, and there are bad actors out there, we can pollute their data and we can pollute it at scale.

• 51:24Casey Newton?

  You know, I, I'm having the, I think, entirely novel feeling of a Hard Fork recording where I'm feeling optimistic about the state of the media. Like, every single time we talk about the internet and the media on this show, I'm like, well, it's going to be really tough, everybody, you know, battening down the hatches, the storm is here. But Matthew, you actually are giving me optimism that, you know, when, when someone with the right incentives, and I do feel like you have the right incentives, shows up with great technology and like the willingness to go out and make a market, maybe by hook and by crook, like we actually still will have a media industry five years from now.

• 51:59Speaker 05

  Well, and I think it, again, and I might lose people because this is, this is starting to get a little, a little woo woo. But everything that's wrong with the world today is ultimately Google's fault. Now, that's a little strong. But Google taught us, they were the first ones to really teach us, that traffic was the deity that we all had to worship too. Now, and again, I think Google has actually been a net positive for the world massively. But Google beget Facebook, which beget TikTok, which is just kind of spiraling down in towards sort of the attention economy of how do I create a cortisol response and get people to click on my stuff so that we can sell ads to them? I think that that's the wrong direction for really furthering humanity.

• 52:44Speaker 05

  If, on the other hand, as we create this, if the giant block of cheese has holes in it, and we can say that if you as a creator go out there and create something that fills in one of those holes and we'll compensate you for that, we'll pay you more for it. What I'm hopeful is that we get a lot more really interesting, long-form, knowledge-generating content, which is what we all want, which was, and the only reason we're not getting it is because all of the incentives are not how do we do great things, but how do we actually just rage bait people into clicking on links? And so I think it's super important that as we go through this, like, it shouldn't just be us figuring this out. And so I've tried to spend as much time with publishers, with AI companies, we've

• 53:27Speaker 05

  working with some of the leading academic economists, others to sort of say, as we think about this market, how do we make sure it's as healthy as possible? Because the dark mirror version of this is not that journalists go away, not that researchers go away, it's that they're all employed by the five big AI companies, right? And so it's not that we're going to go back to kind of the media times of, you know, the early 90s, it's that we're going to go back to the media times of like the Medici's, where, you know, you have just these five powerful families that control all of, you know, academia and research and journalism. And that's that, it's not actually that hard to envision. There will be a conservative one, sort of the Fox News version, there will be the liberal

• 54:12Speaker 05

  one, there will be the Chinese one, there will be the, you know, Europeans attempt at one. And, and those will be the things that are out there and knowledge gets consigned behind that. I don't want that to happen. And they think that the key to making that not happen is figuring out how we can have a healthy market with lots of sellers of content, lots of buyers of content. And make that as robust as possible.

• 54:34Kevin Roose?

  Yeah. Now, Matthew, I know in the past you have expressed some concerns about the power that you and CloudFlare have by virtue of, you know, sitting in front of 20% of the internet, as you say. And when we've talked before, it's been, some of it has been in the context of the decisions you've made around content moderation, basically deciding whether or not to protect websites with extreme or violent content on them. You sort of famously pulled protection from 8chan, which led that site to go down after some mass shootings. I think a lot of people understood that, but you were concerned at the time that this sort of unilateral power that you had to sort of like wake up and

• 55:18Kevin Roose?

  make a big change to the structure of the internet wasn't something that maybe anyone should have. So I'm curious, like, when you were deciding to implement this change, to push the button to block all the AI crawlers by default, did you worry about that exercise of power and how sort of unilateral it was?

• 55:36Speaker 05

  Totally. And I think we take that responsibility really seriously. But what we realized was all the publishers were sitting around saying, we're dying, we're dying, we're dying. And no one was doing anything about it. And so, like, if you're in that situation and you see something that really matters, I mean, the internet really matters, and we should be fighting for it, and we should be protecting it. And it's dying because the business model behind it is dying. And at no step have we said we're the only solution. In fact, we've tried to work with as many different, even competitors as possible to say, this is important, let's do it. And we don't pretend that we have the answer or that it won't evolve over time. But we do know that the first step in any market

• 56:21Speaker 05

  has to be creating scarcity. And so that's what we did on July 1st.

• 56:25Kevin Roose?

  Got it. Well, Matthew, thank you so much for stopping by. Really interesting experiment. I'm excited to follow it and see how it plays out.

• 56:31Speaker 05

  Thank you guys for having me.

• 56:32Kevin Roose?

  Thanks, Matthew.

• 56:37Kevin Roose?

  When we come back, we'll pass the hat for a big round of HatGPT.

• 57:03Speaker 00

  Support for this podcast comes from Dell Technologies. It's time to put AI to work. Is your business ready? The challenge isn't getting started with AI. It's scaling beyond pilots while maintaining control of your data, governance, infrastructure, and costs. Dell AI Factory provides the integrated foundation you need to scale AI securely and efficiently with one trusted partner. Dell, AI on your terms. From idea to impact. Learn more at dell.com/yourwaytoAI. That's dell.com/yourwaytoAI.

• 57:37Speaker 03

  Companies are pouring money into AI, but few leaders can explain how those AI investments work with the systems you already have. You know, the ones that actually run your business? Without systems, you don't have to start over with AI. You can choose the projects where AI will have the most impact. And Outsystems gives you the platform to make it agentic on your terms with complete control every step of the way. So you can trust AI with the mission critical systems that run your business. Own your agentic future. Learn more at outsystems.com/NYT. That's outsystems.com/NYT.

• 58:12Speaker 06

  Ripple has been powering modern finance for a decade. It's not our first trip around the blockchain, and it won't be our last. We're leading the charge to bridge the gap between existing financial systems and the future of finance. You can have it all. Legacy, stability, and digital innovation. Systems that work with your other systems. Because with the right partner, the question isn't crypto or traditional finance. The question is simply, why choose? A new standard of modern finance. It's happening with Ripple. For informational purposes only. Availability of products can vary. May be provided by different Ripple entities.

• 58:46Kevin Roose?

  All right, Casey, it's time to pass the hat. We are playing another round of our favorite game, HatGPT.

• 58:59Kevin Roose?

  ChatGPT, of course, is our game where we pick tech headlines out of a hat. We riff on them and we eventually yell at each other to stop generating.

• 59:07Casey Newton?

  Which they don't even say in ChatGPT anymore, sadly. It's already become a throwback expression.

• 59:12Kevin Roose?

  Yes. So let's use this week our brand new, very pretty Hard Fork hats. Casey, would you like to tell the people where they can get one of these?

• 59:22Casey Newton?

  You can go to nytimes.com/hardforkhat, Kevin. And if you haven't already, become a subscriber to New York Times Audio. And in exchange for that, and a full year of all the great podcasts, including the entire Hard Fork back catalog, we will send you this very hat.

• 59:41Kevin Roose?

  Yes. It will not have slips of paper inside it, though.

• 59:43Casey Newton?

  Or will it? There's only one way to find out.

• 59:46Kevin Roose?

  God, we're merging the same person. It's terrible. Okay, so I'm gonna put the slips in.

• 59:51Casey Newton?

  Okay.

• 59:51Kevin Roose?

  Jiggle them around a little bit there, mix them up. And then why don't you pick the first one?

• 59:56Casey Newton?

  All right.

• 59:58Casey Newton?

  LeBron James lawyer sends cease and desist to AI company making pregnant videos of him. This is from our friend Jason Keebler at 404 Media. He writes, "The creators of an AI tool and Discord community called Interlink AI that allowed people to create AI videos of NBA stars says that it got a cease and desist letter from lawyers representing LeBron James. AI-generated videos of James, Kevin, included scenes like James as a homeless person, James on his knees with his tongue out, and James lying on a couch clutching a pregnant belly." So I guess my first question here is, who is LeBron James? Just kidding. He's a very famous basketball player. Kevin, what do you make of this?

• 1:00:38Kevin Roose?

  So this is obviously going to be a thing for celebrities. They do not like their names and likenesses being used without their permission. And especially if what you're doing with LeBron James's name and likeness is turning him into a pregnant person. Did you watch any of these videos?

• 1:00:55Casey Newton?

  I did see a couple of them.

• 1:00:57Kevin Roose?

  They're very disturbing.

• 1:00:59Casey Newton?

  Some of them, I think, are like on the end of the spectrum that is just like sort of like surreal and funny. And then some of it is like... Also just like racist and horrible. And regardless of all of that, it's clear that LeBron James did not give permission for this. And it's an interesting story because as far as we know, this is one of the first known times that a celebrity has objected to the misuse of their likeness by an AI company.

• 1:01:24Kevin Roose?

  Yeah, and I am worried that it seems like they have automated the jobs of the M-Prag community. M-Prag is, is of course.

• 1:01:32Casey Newton?

  Yeah, where is that? I've been meaning to ask you about that.

• 1:01:35Kevin Roose?

  It's a niche fanfiction thing that for years, people have been sort of creating these like animated fanfiction cartoons of like Sonic the Hedgehog becoming pregnant. Why are they doing this? I don't know. Couldn't tell you. Not an M-Prager myself. But there were a lot of hardworking M-Prag artists out there who now have been put out of jobs by these AI tools. So for that reason, and that reason alone, I think we should take a hard stand.

• 1:02:03Casey Newton?

  I now want to say retroactively to my parents, do not listen to this segment.

• 1:02:09Kevin Roose?

  And definitely don't Google M-Prag.

• 1:02:11Casey Newton?

  All right, stop generating.

• 1:02:12Kevin Roose?

  Okay, next up.

• 1:02:17Kevin Roose?

  Oh, this is a fun one. Sam Altman warns there's no legal confidentiality when using ChatGPT as a therapist. This one comes from TechCrunch. Sam Altman, the CEO of OpenAI, went on the popular podcast by Theo Von last week, where he acknowledged that OpenAI might be legally required to produce a user's conversations with ChatGPT in the case of a lawsuit. Now, I saw this clip going around and I thought Sam Altman continues his terror campaign against the Hard Fork Podcast. As you will remember, at our live show, he burst onto stage and then peppered us with questions about this lawsuit between OpenAI and the New York Times,

• 1:02:57Kevin Roose?

  which has resulted in OpenAI having to retain conversations between ChatGPT and its users. He's very upset about this, doesn't want this kind of document retention to be required, and has started advocating for the privacy guarantees that AI companies.

• 1:03:17Kevin Roose?

  should be allowed to make with their users.

• 1:03:19Casey Newton?

  Look, this one is important because a lot of people already are using these chatbots as therapists. They're having therapy style conversations. And I think most people are not thinking a lot about what is happening to that data. Some companies may want to erase that data for, you know, user protection reasons. Other companies might want to keep it forever and create a detailed profile about you. And then like rent some information to advertisers. So I would love to see some kind of legal intervention, regulatory intervention come down and say, you're allowed to use information submitted to chatbots in these ways. You know, users should have a full view into what a chatbot knows about them. What kind of information is being stored about them? They should be able to delete it, right?

• 1:04:04Casey Newton?

  We just need a lot of like data and privacy stuff around this sort of thing. And so I have to say, I'm like, I'm grateful for Sam Altman for at least saying like, hey, by the way, you don't have legal protections here.

• 1:04:15Kevin Roose?

  Yes, we are allowed to use data however we want in training our models. But God forbid, you know, anyone else want to use data about our users after the fact.

• 1:04:24Casey Newton?

  All right, stop generating. All right, your turn.

• 1:04:29Casey Newton?

  How to catch a wily poacher in a sting. A thermal robotic deer. This is from James Finelli at the Wall Street Journal. Wildlife enforcement officers turn to a Wisconsin taxidermist to make remote controlled robots that look like wild animals to catch poachers, Kevin. To make his decoys, a man named Brian Wolslegel applies the skin of a dead animal to a mold made out of polyurethane. He affixes glass eyes and plastic ears. The circuitry to make the decoys move comes from parts for remote controlled cars. With some AA batteries, officers can remotely operate the Bambi bots. Kevin, would you be interested in one of these to ward off some of the poachers on your property?

• 1:05:11Kevin Roose?

  Yeah, I think there's been a lot of poaching attempts made against my livestock, and I won't be standing for it. So I'll be, I'll be buying one of these.

• 1:05:17Casey Newton?

  Here's, here's where this man's real opportunity is. He needs to use the same technology to build decoy AI researchers and put them in OpenAI so that when Mark Zuckerberg comes onto the property, he mistakenly poaches one of the robots instead of one of the human researchers. So if I'm Sam Altman, this is what I'm doing.

• 1:05:36Kevin Roose?

  Yeah. In general, I think like robot animals, I'm not a big fan of. Like, there's actually, on my block, there's a guy with a robot dog. He runs like a STEM program for kids, and I've seen him like walking his robot dog out on the street. It's very disconcerting. Let's just say.

• 1:05:54Casey Newton?

  Top 10 signs you live in the Bay Area, by the way.

• 1:05:58Kevin Roose?

  Yeah, let's just say these things are not entirely life-y like yet, and they do still give you the willies.

• 1:06:04Casey Newton?

  All right, stop generating.

• 1:06:07Kevin Roose?

  Next up, did a guy just save a picture of a bird to a bird's brain? This one comes to us from Sean Hollister at The Verge. It's about a YouTuber named Ben Jordan, who has a popular channel about music and acoustic science, who was able to get a bird, a starling, to reproduce a spectrogram image in sound. Casey, did you hear about this story?

• 1:06:29Casey Newton?

  I, this is actually my favorite story of the week.

• 1:06:32Kevin Roose?

  What happened?

• 1:06:32Casey Newton?

  So, if I get this straight, the YouTuber converts an image into a sound. He then plays the sound for a bird. The bird then repeats the sound. And so, in that way, you can say that the YouTuber was able to store an image in the bird's brain. He used the bird as a data transfer device, like it was a, like a disk drive.

• 1:06:58Kevin Roose?

  Yeah, so you almost got it right. Because what happened is he takes this drawing of a bird, he converts it into a spectrogram, then he plays that sound for the bird, the bird mimics it back, and then he takes that recording and turns it back into a spectrogram. So he's essentially going image to sound, transfers the sound to the bird, bird transfers it back, converts it back into an image. Now, Casey- Why would you do something like this?

• 1:07:21Casey Newton?

  Because you have a YouTube channel and you're trying to get a lot of subscribers. Now, how close was the image, the second image to the first image?

• 1:07:29Kevin Roose?

  Apparently it was quite close. This was like, you know, the bird is a little bit of a lossy compression device, but you actually were able to sort of like recognize it as a line drawing of a bird after the fact, which is kind of amazing.

• 1:07:42Casey Newton?

  Honestly, hats off to this person. What a bizarre idea, but fantastic.

• 1:07:48Kevin Roose?

  You know, I uploaded an episode of our podcast to a bird.

• 1:07:51Casey Newton?

  Oh, yeah?

• 1:07:51Kevin Roose?

  Yeah, it's called Hard Stork.

• 1:07:54Casey Newton?

  Okay, sir. Why not? Stop generating.

• 1:08:00Kevin Roose?

  All right, you're up.

• 1:08:01Casey Newton?

  Okay.

• 1:08:05Casey Newton?

  Meta is going to let job candidates use AI during coding tests. This is from Jason Keebler at 404 Media again. Meta told employees that it is going to allow some coding job candidates to use an AI assistant during the interview process, according to internal Meta Communications. On an internal message board for the company, there was a post that called for mock candidates, because apparently they're going to let employees, you know, do one of these sort of AI assistant interviews so they can work out all of the kinks. So it looks like Meta is going all in on using AI coding agents to write code, and also is just not going to try to stop people from cheating on their job entrance exams anymore.

• 1:08:40Kevin Roose?

  This is total Roy Lee victory.

• 1:08:42Casey Newton?

  Absolutely.

• 1:08:43Kevin Roose?

  So this is like a vindication of what Roy told us when he came on the show several months ago, which is that these leap code interviews are totally cooked, because now people can just use these tools like the ones Roy Lee is developing to cheat on their interviews. I guess Meta is sort of seeing the writing on the wall and saying, you know what? Go ahead, use your AI. We'll design our new test, which I think is probably a good outcome. What do you think?

• 1:09:07Casey Newton?

  Yeah, I mean, I think I'm interested to see how this affects the candidates that they attract, the quality of the engineers that they can recruit. You know, I am ultimately persuaded that people are going to be using these tools in the workplace anyway, so why not use them when you're doing the actual test to get the job?

• 1:09:24Kevin Roose?

  Now, do you think they make the people who are making a billion dollars a year take the tests when they come in?

• 1:09:30Casey Newton?

  Yeah, they just give them the really hard version. Say if you do a good job, we'll give you a billion dollars.

• 1:09:34Kevin Roose?

  God, what a weird thing. Can you imagine being on boarded to a new job? And like you go through the training and it's like, you know, here's your benefits package and you're just sitting there thinking I'm making a million, a billion dollars.

• 1:09:46Casey Newton?

  Yeah, they're like, do you want to put anything in your, your health savings flex account this year? What about the commuter benefit? Do you want the 40 bucks for Bart this, this month?

• 1:09:58Kevin Roose?

  You're like, I'm making a billion dollars, people. I'm not sitting through the IT training.

• 1:10:05Casey Newton?

  Okay.

• 1:10:06Kevin Roose?

  Okay.

• 1:10:07Casey Newton?

  Stop generating.

• 1:10:08Kevin Roose?

  Stop generating.

• 1:10:10Casey Newton?

  Here's one, Kevin. This is from Taylor Lorenz at User Mag. Substack sent a push alert encouraging users to subscribe to a Nazi newsletter that claimed Jewish people are a sickness and that we must eradicate minorities to build a white homeland.

• 1:10:23Kevin Roose?

  Oh, boy.

• 1:10:24Casey Newton?

  Yeah. And you know, I have to say, Kevin, rarely have I felt so smug in my entire life as I did when I read this story. Long time listeners of the Hard Fork Show may know that I moved Platformer off of Substack last year after some other folks had found a bunch of pro-Nazi websites on the network and Substack would not commit to doing any proactive searching for these blogs to get rid of them. And the main reason that I wanted to leave was I thought these people are building amplification features and inevitably they're going to just start promoting these things and it might be unwitting and it might be intentional, but either way, I don't want any part of it. And so now, sure enough, a bunch of people who had the Substack app installed yesterday just got a ringing endorsement for a Nazi blog.

• 1:11:07Kevin Roose?

  Oh, boy.

• 1:11:07Casey Newton?

  Yeah. Now, Substack did basically say that this was a huge mistake and they took the offending recommendation system offline and they're going to rejigger it so it doesn't happen again. But, well, it, it did happen and I feel quite vindicated.

• 1:11:21Kevin Roose?

  They did not see it coming.

• 1:11:23Casey Newton?

  Yeah, that's, that's a good way of putting it.

• 1:11:25Kevin Roose?

  Yeah.

• 1:11:26Casey Newton?

  All right, stop generating.

• 1:11:28Kevin Roose?

  All right, last one. Why Amazon wants an AI bracelet that records everything you say. This comes from Nicole Wen at the Wall Street Journal. Amazon is acquiring a company called Bee. Bee makes a wearable device, B E E, that transcribes all the conversations in your day, including when you talk to yourself. It then uses AI to turn that giant word soup into a searchable history, offering up key events and even to-do lists based on your chatter. Friend of the pod and Wall Street Journal reporter Joanna Stern reported on her own experience testing out the Bee bracelet earlier this year. She described it as impressively useful and also, quote, "really fucking creepy." Casey, what do you make of this bracelet? And will you be buying one?

• 1:12:13Casey Newton?

  I'm not gonna buy one for myself. You know, generally, I don't want a detailed record of everything that I say during the day. You know, one of the reasons why we podcast is so that most of what I say can be edited out. You know, so the idea that I would just sort of have this unfiltered record, you know, stored in an AWS bucket doesn't really appeal to me. Also, as I read the reviews of these, uh, these devices from Joanna and others, nobody really seemed like they were getting a lot of value out of it. It was like, oh, yeah, I like told my husband I should get milk. And like, now I get an email that's like, hey, remember you want milk. It's like, is that really worth, you know, giving up all of your privacy in perpetuity?

• 1:12:51Kevin Roose?

  Well, and all the privacy of everyone that you talk to.

• 1:12:53Casey Newton?

  Yes, exactly.

• 1:12:55Kevin Roose?

  Like, that, that is the, the worst part of this. Like, I have, at times, been recording interviews and like accidentally left the voice memo running for like an hour afterwards. And it's never that it's that interesting, but sometimes it does catch other people's conversations in there and then I feel bad about it and delete it. But with the Bee bracelet, this is the whole point. It's just logging you all the time. I don't think people are gonna be that excited about that.

• 1:13:18Casey Newton?

  Well, here's what I'm looking for is, uh, for keeping the Bee on you at all times to be a condition of continuing to work at the Washington Post, owned by Amazon founder Jeff Bezos. Because they're going through a lot of turmoil right now and there's a lot of people who are like leaking stuff to the media. So I think it's gonna be like, hey, uh, we, we need to check your Bee. What have they been, what have they been saying about us?

• 1:13:39Kevin Roose?

  So it sounds like you are not going to be on the early beta tester list for the Amazon bracelet. Uh, maybe I will.

• 1:13:45Casey Newton?

  I'll be minding my own beeswax.

• 1:13:48Kevin Roose?

  Stop generating.

• 1:13:49Casey Newton?

  Okay.

• 1:13:50Kevin Roose?

  And that's HATGPT.

• 1:13:51Casey Newton?

  Yay.

• 1:13:52Kevin Roose?

  Thanks for playing.

• 1:13:53Casey Newton?

  We won again.

• 1:13:56Kevin Roose?

  What's that they used to say on Whose Line Is It Anyway? Where everything's made up and the points don't matter?

• 1:14:00Casey Newton?

  That's right. That's GPT very similar.

• 1:14:02Kevin Roose?

  You know, my three-year-old is very obsessed with winning and losing.

• 1:14:04Casey Newton?

  Oh, yeah?

• 1:14:05Kevin Roose?

  Yeah, every time we do anything, he says, "I won, you lost." So I'm gonna start doing that with you.

• 1:14:29Speaker 00

  Support for this podcast comes from Dell Technologies. It's time to put AI to work. Is your business ready? The challenge isn't getting started with AI. It's scaling beyond pilots while maintaining control of your data, governance, infrastructure, and costs. Dell AI Factory provides the integrated foundation you need to scale AI securely and efficiently with one trusted partner. Dell, AI on your terms. From idea to impact. Learn more at dell.com/yourwaytoAI. That's dell.com/yourwaytoAI.

• 1:15:03Speaker 03

  It's time to innovate at the speed of AI without compromising control. That's Outsystems. Take a look at what happened with one global retailer during a period of massive expansion. They used Outsystems to build AI agents to automate invoicing and translation processes, shifting 95% of their daily store operations to a new digital app, saving thousands of hours of manual work. Outsystems gives you the flexibility to deploy how you want and the governance to do it safely. Own your agentic future. Learn more at outsystems.com/NYT. That's outsystems.com/NYT.

• 1:15:38Speaker 06

  Ripple has been powering modern finance for a decade. It's not our first trip around the blockchain, and it won't be our last. We're leading the charge to bridge the gap between existing financial systems and the future of finance. You can have it all: legacy, stability, and digital innovation. Systems that work with your other systems. Because with the right partner, the question isn't crypto or traditional finance. The question is simply, why choose? A new standard of modern finance. It's happening with Ripple. For informational purposes only. Availability of products can vary and may be provided by different Ripple entities.

• 1:16:14Kevin Roose?

  Hard Fork is produced by Rachel Cohn and Whitney Jones. We're edited by Jen Poyant. We're fact-checked by Caitlin Love. Today's show was engineered by Chris Wood. Original music by Alicia Baty-Tup, Rowan Nemusdo, Alyssa Moxley, and Dan Powell. Video production by Sawyer Roquet, Pat Gunther, Jake Nickel, and Chris Schott. You can watch this full episode on YouTube at youtube.com/hardfork. Special thanks to Paula Schuman, Qui Wing Tam, Dahlia Hadad, and Jeffrey Miranda. You can email us, as always, at [email protected]. Send us all the images that you've uploaded to Bird's Brains.

• 1:17:13Speaker 04

  AI isn't just opening new business opportunities, it's changing the way businesses get work done, the questions they ask, the decisions they make, and the routine tasks they can hand off. And nobody understands that better than NetSuite. That's why they put together a free expert guide you need to grab today, aligning for the agentic era how AI is changing everyday work. Download it and you'll see exactly how AI is reshaping work across every part of your organization and how to make sure you come out ahead. Get your free guide at netsuite.com/NYT. netsuite.com/NYT.