Welcome to the Vergecast, the flagship podcast of Mythos class models. I'm your friend David Pierce, and today on the show, we're going to talk about AI. We got a lot of questions from all of you over the last couple of weeks after Google IO and Microsoft Build and Apple's WWDC about all things AI. And then we had a just truly chaotic blowup between Anthropic and the US government over the new model Claude Fable and whether it is the most dangerous thing anyone has ever invented in history. So we're going to get The Verge's senior reporter Hayden Field. She's going to come and we're going to answer a bunch of your questions about all things AI.
But first, here's a look at everything else happening on The Verge today. This is 90 seconds on The Verge for Tuesday, June 16th, 2026. Snap's first consumer-grade AR smart glasses are finally actually apparently coming out this fall. That's what the company announced today. They're called Specs. They're thick and wide and black and they kind of look like spy glasses that James Bond would have worn in the 1990s. I don't know if I mean that as a compliment or an insult, but there it is. But the best part is they're going to cost $2,195.
It's been a full decade since Snap's first ever pair of spectacles went on sale. Back then, they were only $130, and they weren't a huge hit. Even then, I have no doubt that the tech here will be impressive, including the built-in display and the new chips, but this one might be a steep climb for Snap.
Meanwhile, Microsoft launched new Surface devices, a new Surface Laptop and a new Surface Pro, both of which are powered by another Qualcomm chip, the Snapdragon X2. They look good. The spec sheets are great. The colors are solid as always. And unfortunately, as with basically every gadget right now, they are too expensive. The Pro starts at $1,500 and the laptop at $1,600. Both up a hundred bucks from the last generation. I am still excited about them, though. The Surface lineup is some of the best stuff going in Windows World and has been for a long time. And as these Qualcomm chips get better, so do these laptops. Microsoft, in particular, is making really big battery life claims, and I hope they pan out.
And finally, happy birthday to the Trump phone, which was first announced a year ago today. Back then, the Trump phone was made in America and was going to be shipping last September. Today, it's an obvious dupe of an HTC phone that is very much made in China, and as far as I can tell, it has only shipped to a bunch of YouTubers and tech reporters. I suspect most people at this point have just written off their $100 deposit. But hey, maybe year two will be a big year for Trump Mobile. I would not hold my breath. You can read more about all of this at theverge.com. That is 90 seconds on The Verge for June 16th.
All right, let's answer some questions about AI. Hayden Field is here. Hi, Hayden.
You are just deep, deep in Anthropic nonsense as you always are. This seems to be like half your beat now is various Anthropic nonsense.
It's true. We have a joke that like anytime news happens on another company, they text me immediately. As in me and Addie, my editor, had that joke because it's like every time something else happens, like, 'Oh, Anthropic's got to get in there.' You know, it's just crazy. Um, but no problem to them. It's just a hilarious coincidence every single time. But yeah, last night we were up till midnight. We got a piece out. Um, there's some crazy battles going on again between Anthropic and the Trump administration.
This is just what we do here. So, I have a handful of questions for you from our audience about all things AI, but everybody's asking us to make sense of what's going on with Mythos and Fable and Anthropic and the US government. So, let me just start with a question from Samuel, which says, 'Can we have a complete Vergecast episode on how this is going to impact the way the world looks at US tech companies and the risk this behavior will cause? What's next? Stopping PowerPoint for the rest of the world?' I like this question because I think it is a good framing for what's happening between Anthropic and the US government and kind of the way the rest of the tech industry and the rest of the world is going to take whatever the outcome is of this story. But really quick, just for people who have not been following this, you did sort of a big TikTok story of what's been happening over the course of the last several days. Just give me the cliffnotes version of what has happened between Anthropic and the US government in the last week.
Yes, that's a great question because the last like five days have really been the crux of the whole thing. So on Friday night when everyone was getting ready for the big games that were happening this weekend, Anthropic was having the opposite experience where they, at like 1:00 p.m., got a call from the US government saying you have 90 minutes to shut down access to Fable 5, its new model that had been hyping for a week, and Mythos 5, its other new model that had been hyping for a week. So, Fable 5 is kind of the safeguarded version of Mythos. They're built on the same underlying model. Mythos 5 is the one that they gave access to enterprise, governments, you know, select partners.
Mythos is the one that they were like, 'If we release this to the world, it will destroy everything. We can't possibly let you have it.' I'm barely overstating what Anthropic said when they put this out.
Exactly. And that's what they still say. And then Fable is the one where it's built on the same exact model, but it has a lot of safeguards apparently, so it's safe to release to the public. Now, we don't know for sure, but the independent red teamers I have spoken with have been pretty impressed with the safeguards so far, which is really saying something. I wasn't expecting that. Now, what happened on Friday was at 1 p.m., yep, they got a call from the US government saying, 'Hey, we are aware of a jailbreak here for Fable, the safe one for the public. So, you've got to shut it down within 90 minutes.' Anthropic asked for more details. They said, 'Okay, well, what's going on?' Like, you know, let's talk about this. And the US government apparently was not happy that they asked for additional details. And they said, 'Nope, it's been more than 90 minutes. You got to shut it down.' And they were hit with an export control directive at 5:21 p.m. on Friday. So that was basically saying, 'Hey, you cannot allow any foreign national to use either of these models even if they work at Anthropic. So if you're not a US citizen, you can't use the model.' And obviously Anthropic was like, 'Well, we can't control that. So let's just shut it down. We can't really — I mean, these are out there now, you know?' So they had to completely sideline them. The whole weekend was spent, you know, trying to kind of make nice with the Trump administration. Apparently, Dario was not available for the first hour that the Trump administration called. Anthropic called the Trump administration back within 15 minutes when that first call came in, but Dario, it took an hour and 15 minutes from that first call. He was unavailable originally. And so that created a whole point of contention as well. There was some rumors that he was on a wellness retreat. I can confirm that he was not on a wellness retreat, but he still was not available for an hour and 15 minutes. So, yeah, there was a lot of drama going on, lots of virtual meetings all over the weekend. Um, then Anthropic ended up flying out three of its senior staff to DC on Sunday. Notably, Dario I don't think was among them. Uh, he was still meeting virtually. And so the in-person meetings on Monday concluded without any resolution. So right now it's still banned. So we'll see what happens. Right now, Anthropic is trying to talk to the Trump administration and say, 'Hey, we saw the jailbreak you're aware of.' And it's actually something that OpenAI's GPT-5.5 can also do. It's not that crazy. So, there, you know, they went from hyping this model for a whole week to saying, 'Oh, well, it's actually not as powerful as you think.' So, there's a lot of interesting things going on here. Amazon is also involved because apparently Andy Jassy was the one to sound one of the first alarms about this to the Trump administration, which is fascinating for a whole bunch of reasons we don't have time for here. But like Amazon's weird role in all of this as a company with like deep ties to everyone in the AI space. Very strange.
It seems to me that there's been a lot of discussion about all of the things that might be happening here. There was a like maybe China hacked it story that I think has turned out to be pretty thin. There's a what if this is just fundamentally about the ongoing relationship problems between Anthropic and the US government. There's a whole thing like you said about are these guardrails completely passable and you can get access to this whole terrifying Mythos system. It seems to me, and I should say this with the preface that there is still an awful lot we don't know and there's a lot of this left to come as with this entire story. But it seems to me that the sort of Occam's razor explanation of this, the simplest possible explanation here is that this is just once again the Trump administration trying to punish Anthropic in particular. And you think about what the Trump administration responds to, which is people in a position like Dario Amodei coming and, you know, bowing at the knees of the Trump administration. That is, you show up to the octagon for the UFC fight and Trump gives you what you want. Like this is how it works. And so Anthropic has largely not played this game. And you can just see how an administration that operates this way would respond by, well, we have to continue to find ways to punish them because they are the enemy. Again, I don't want to say that is the answer, but it seems to me that that is the simplest answer to a lot of this right now. Is that your read too, or are you seeing something different?
I think it's going to be really interesting to see what else they try to squeeze them on. Um, you know, like you mentioned, this is kind of following a lot of months and months of discord between the two, and there's even an ongoing lawsuit right now about the supply chain risk for the DoD stuff. So, I think what's going to be interesting is what else they squeeze them on. Are they going to, you know, make them, you know, bend the knee for DoD stuff to get this out of the export control directive? We don't know. I mean, I think that it's a combination of the two things, what you said and the fact that the government doesn't know that much about this technology a lot of the time. And so if Amazon researchers or Andy Jassy himself are calling, you know, Scott Bessent and saying, 'Hey, I found a jailbreak in the most powerful model that Anthropic has spent, you know, months saying is so powerful it could be a cyber weapon,' you know, they're going to freak out. Um, but what I — so I think that was the initial situation. It's the fact that they don't know that much about this technology. If someone who does know a lot about it is telling them it's a problem, they're going to freak out. But then they try to talk to Anthropic. Anthropic says, 'Okay, let's discuss this. Let's like, you know, here's some more information. We'll come to a conclusion.' And they immediately said, 'No, shut it down.' I think that's the part that was probably a result of the administration's current communications with Anthropic. Apparently, a lot of reports have come out saying that they're just not good at talking to the administration. They don't, you know, communicate in the same language. They, aka, they're not good at like making nice. And I think that that is probably the reason this kind of blew up in their face, especially with the foreign nationals thing. I mean, everyone in the tech industry is just absolutely aghast at that.
Yeah. So let's go back to Samuel's question at the beginning here, which is about what would this mean for other US tech companies and others looking at building tech like this. Because what you have here is the US government going to, I would say, the leading AI lab in America, if not in the world, and saying the best model you've ever made, which might be the best model anyone has ever made, you are not allowed to release. And whether or not that is actually like a thing the government should be and is able to do is sort of a separate question. It did do it, right. And so you talked to a bunch of people for this story about what the sort of downstream effects of something like this might be if this continues to play out the way it is playing out. What do you think other folks in tech and around the world are looking at in this, and what kind of changes are they going to make in how they think about this technology?
Yes, that's a great question. It is crazy out there right now, and I think whenever there's a lot of uncertainty in an industry, it's just not good for business, and this is like peak that. So, um, I talked to a lot of people for the story. Yes. And it's interesting because people in the cybersecurity industry and the tech industry, a lot of leaders have actually signed a public letter saying basically that they think this is really stupid, even if they are not, um, even if they don't agree that regulation should be a thing. They literally wrote, 'We don't all agree on that, that there should be regulation at all, but if there is regulation, it certainly shouldn't be this.' Interesting. Um, and so yeah, I mean, uh, one guy I interviewed, Alex Stamos, who put that letter together, he said that the industry right now is like awash with backup contracts being signed with non-US companies, and open-weight models are being deployed on alternative hardware arrangements because, you know, political risk is now a huge part of companies' business plans in a way they didn't want it to be before. Um, he said they're laughing at us in Beijing right now. One of America's champions is being kneecapped by the US government while we're in a race with the Chinese. It's just incredibly stupid. That's why I wrote the letter, and I think that's why a lot of people signed on to it. And yeah, a lot of the people I spoke with echoed that sentiment. It's like, um, we're already, you know, the general consensus in the industry right now is that we're about six months ahead of China in AI. If that's like the best, seems to be like the hopeful scenario for how far ahead we are in AI. Best case scenario, six months is kind of the consensus. And you know, this is just seen as being like a very unnecessary, strange, kneecapping-ourselves moment.
There are a lot of ways that this mirrors what was going on with the Pentagon deals, like you were talking about, and we talked about on the show, that the sort of moment where any minute now OpenAI is going to come out with some announcement about how it also launched a cyber weapon of a model and everything's going to be great. Um, there are weird machinations left to be done here. So we should come back to this. But I think this question of the race to AI supremacy in the world next to the government trying its best to cut off at the knees the leading American AI company at this moment is just very strange. And I feel like Alex Stamos in particular, by the way, is somebody who knows, right? Like he was the chief security officer at Facebook for a long time. He ran the Internet Observatory at Stanford. He's like, he is as good a cybersecurity guy in this space as there has been. Uh, so to see him turn this quickly on this idea was pretty fascinating to me.
Exactly. And the other crazy thing to me is that OpenAI, um, you know, and a bunch of competitors actually already have cybersecurity-focused models that they say are as good as Mythos. Like the consensus is that they aren't, they're still behind, but they're marketing them as a similar product. And so if Anthropic undergoes this, how long until, um, its competitors have this same type of thing happen if this isn't resolved? And also the foreign nationals saying, 'Okay, are all the US tech companies going to only allow US citizens to work on the products?' It's just, it's a huge mess. And I think hopefully we'll see this resolved in the next couple days, but I mean, even the fact it's taken this long is really surprising to a lot of people.
