THE US GOVERNMENT JUST KILLED THE WORLD'S BEST AI

The US government ripped away the world's best AI model and nobody's entirely sure why. Classic.

Anthropic released Claude Fable 5, let everyone play with it for three glorious days, then got a phone call from the Commerce Department at teatime on a Friday telling them to switch it off for the entire planet. 

National security, apparently. The same government that blacklisted Anthropic for refusing to let the military use Claude for autonomous weapons now says the model is too dangerous for foreigners. 

Which is everyone. Including the foreigners who work at Anthropic. You genuinely could not write this.

00:00 The ban lands
01:15 Hull celebrates promotion with blue drinks
02:00 What Fable 5 actually did
03:30 Can you VPN around a government order?
05:00 Best marketing Anthropic never paid for
07:00 The Pentagon grudge theory
08:30 Your own engineers locked out of your own tech
10:00 Building your own AI in the shed

#AIban #ClaudeFable5 #AIsovereignty

Subscribe to our channel on YouTube

Check out our website at temporallyscripted.com

Follow us on X, too!

Transcript

0:00

The world's most powerful available AI has been given to us and taken away in the last week, Adam. So did you get a chance to use Claude Fable 5 before the US, uh, banned it for any internationals to use? Well, unfortunately not. I was planning to, and then it was literally on the day I was gonna use it as well, 'cause I'd watched some videos showing examples of how, just how much better it was than Opus, and so some of these examples for websites, it just looked really good, and I was like, "Oh, right, that's one good use for, uh, Fable 5. Uh, I'm gonna do that tonight." Fire up Claude, Claude Fable 5 is not available. I was like, "Oh, man." Uh, yeah, so that kind of scuppered my plans, and my website still looks like a, uh, a 12-year-old with serious problems made it. It is interesting how AI slop has transitioned from writing to website development, and quite a lot of people don't notice the difference. There are ways and means of making good websites, but this is not a how-to video, so we're not gonna go into that. I, I think we've entered something that we've never really seen before. Of course, technologically we've seen chips being banned by being used in certain countries. Uh, for those of you watching in Hull, that's chips as in the ones that go in computers, not the ones you have for dinner with your blue drinks. Um, but yeah, it's weird. I've never seen a- an AI be brought out and be taken away or, or a piece of technology like this that's not hardware, right? This is software usage, whatever else, and the people of Hull will accept that 'cause they just got promoted to the Premier League, so they, you know, they've got things to be happy about and- So celebrate with a, a trip to the chip shop or something. Get two blue drinks. Exactly. I don't know. Two blue drinks for a pound. Um, yeah. I, I did get to use it, but not massively. Uh, it was certainly a stronger model. It was very, very expensive, which I think is what we're gonna see with AI anyway. Things are gonna move forward and be more expensive. And actually, you know, Mythos or Meethos or however you wanna say it, uh, was never really available anyway. It was only available to people that are certain partners of Anthropic to use. It wasn't actually- Wasn't there like, um, they had to have some, like, level of security or something as well, or it was a certain use case for it? And I, I remember- Yeah … reading it was something to do in like, uh, finance, uh, companies. Yeah, and also I think you could-- If it, it-- I saw a message when I was using Fable that said, "If you're gonna use it for computer security related things, it will automatically switch back to Opus." So there was definitely- Hmm … some concerns about it being able to find… 'Cause remember, it could find the holes in security systems and data systems and all of these kind of things. So there was like- Yeah … a, a massive concern in that area. But yeah, it's, um, it's unprecedented and I wonder what they know about the usage that we don't know. Um, yeah, completely. I mean, like you say, it's the first time we've, we've had, uh, something like this, and I guess with the AI, just like in the past someone would say, "Okay, I'm writing a novel and I need to know how to do blah, blah, blah, blah," that a terrorist might do, and it goes, "Oh, actually if you're writing a novel, that's fine. Here you go." Um, and I guess in, in maybe not quite a similar way, but something that's o- on a, a certain level you could say to it, "Okay, c- can you find the security flaws in, in my program?" Uh, but that program could be something that some other companies also use. I don't know, I think there would be various ways that people could not, not completely jailbreak it, but maybe use it for nefarious purposes. Yeah. And I, and I wonder, I wonder if anybody's… I guess that they're, they're smarter than this, but I wonder if anybody's tried to access by VPN yet, and if that just bypasses it or if it's slightly more technologically secure than that. Yeah. Because a lot of the times I've seen in the past it's like, oh, ChatGPT, whatever's come out, and it's only available to users in the United States, and, and I fire up my VPN, uh, which I won't mention the name of because we don't have a, a deal with them yet. Yet. Um, and, and I fire up my VPN, which is the main reason I use my VPN, and I can just use whatever thing is available in the US normally. I've not actually gone to the trouble of, of testing whether or not my Anthropic package works that way, but I wouldn't be surprised. I know that, um… Yeah, I know a couple of years ago I did that with Google. Um, I can't remember, it was some AI things at Google Music, spelt in a funny way, I can't remember. But it was some kind of AI music generation, and it was like, "Oh, that's cool," and it… "Oh, not available here." It's like, "Okay, well, VPN. Oh, right, I can use it now. I'm, apparently I'm in the USA." Um- Mm-hmm … but yeah, it could be something like that. Do you think this is, um, do you think this is bad or good publicity for Anthropic? I think it's great. Yeah. As, as they say, all publicity is good publicity. I, I think the, the question is gonna be how much custom they lose to models from other countries because They're almost acting like they're the only company to have this level of tech, right? This level of advancement in AI models and capabilities. That's the, the right word, capability. But like the Chinese models are working on similar capability. There's models from other countries available that have got similar capability, and now obviously people will chase the best working AI for their needs, right? So if suddenly one of the companies in China releases a version that's similar in capability to Fable, they get first usage advantage, don't they?'Cause like one of the reasons you don't wanna change from ChatGPT to Claude, for example, if you've always been a user of ChatGPT, is it's got all your chats and all your history and all your memory, and you're used to the interface, and it's like it's human nature. We'd rather stay with the thing that we're familiar and comfortable with, unless it gets to a point where it's just like really damaging us to stay, right? We, we need-- Humans are, are pushed by fear and discomfort rather than, you know, desire usually. It's like- Yeah. That's mankind … you move because someone's, yeah, sticking a knife in your back and making you move. If you can avoid it, then, then you avoid it. So I, I think it gives other companies and other countries that first-mover advantage if it's not- Yeah … lifted. I, I, I think some of it is also vendetta of the American government against Anthropic because they refused their terms on being their partner for military stuff, so they went to ChatGPT and OpenAI instead. Oh, right, which would make complete sense."You won't help us with this, so we're not gonna help you with that. See you later." Um, but at the same time, if, if that's, if that is what the American government were doing, I think it might kind of backfire in some ways because I think, I think really, uh, Fable, uh, getting taken down is actually a really good thing for Anthropic because it, it's just a great bit of marketing for people like,"Oh, wow, it's so dangerous that it got taken down by the US government. I want it." You know? And this comes just at, at a time when, uh, they're talking about going public. I don't know if a date's been set yet, but they're gonna go public, sell shares, do a massive round of, uh, money raising. And so it could actually be a really good time, and it, it seems like really good publicity. Yeah. Yeah. And I wonder what it means for their, their engineers and their people and how they're gonna work on that, because it's not just foreign nationals i- in the general public, it's foreign nationals even within Anthropic themselves that are not allowed to use this tech anymore. Huh. That's crazy. Yeah. So I don't know how they are separating that or working that through or even how it's being monitored. Uh, it's, you know, unprecedented. And then if the, you know, the government has a relationship in terms of military and national security with OpenAI, where does that stand if they go and find out all of the ins and outs and workings of Anthropic and who's using what and how they're using it? Are they-- The- is that information gonna get passed on to the company that is working on national security, or is there like… How does that work? Because surely- Yeah … if you're saying,"Well, it's national security," you're allowed to take whatever and provide it to whoever in order to- Oh, I see what you mean increase national security. Yeah. Like- So can they just like go in there and say, "Uh, as a matter of national security, we need all of your code"? Who, who knows, right? So- Yeah … yeah, it's crazy. I mean, it's, it's such a level that it's, it's hitting now for it to be at this stage. You know, I mean, i- in the past there's been, I don't know, I guess things about people generating certain images or whatever, but we've never with AI, I don't think, got to a level where there's been security concerns of this kind of level, and it just shows like the, the actual capabilities of these things are just really getting advanced- Mm. Yeah … to, to what seems like a really crazy level. Yeah, I think it will also vastly accelerate people building their own AI models and server farms and capabilities where they don't have to be connected to any kind of network or, or anything like that. Like I get asked this quite a bit by clients already is like, you know, "Is there a way of me giving information and data that won't go to the cloud?" And it's like, well, yeah, you need to download your own model like Gemma 4 or something from Google, and then you can have your own AI running from your system, right? Granted, your own system is not gonna be the most powerful, but yeah. W- when… how quickly then do we get to the stage where people start sharing bigger systems and investing in them as a group to build a totally offline version of these models to do whatever they see fit? Yeah, 'cause that's what, um, that's what caused Apple to run out of Mac Minis, wasn't it? But Mac Minis were just ideal for this kind of thing 'cause it's like the silicon chips. Mm. They run really quick. And yeah, you… they were really good for just installing your own AI on it and just not having anything else on there, just the AI doing its thing and doing tasks for you. One of the only electronic items in history that is now more expensive secondhand than it was bought new. Yeah, which is mad, isn't it? Mm. This is, uh… yeah, what a world that we live in. So yeah, I mean, I think looking forward w- we will see quite quickly how, you know, how this unfolds, whether or not ChatGPT release something, because they seem to be on this increase in releasing models curve again, which we've seen a few times over the last few years. But I wonder if now this will be a, a suppression until we see what's, what's gonna happen. But it's certainly the first time we've seen models just be held in one country and also held to a certain group of people, which is what we've seen with MiFast anyway. So definitely an interesting one. Yeah. I'm sure there'll be more stories that we cover here on Temporally Scripted, and if you wanna come and listen to those, you give us a like, a subscribe, a comment. Let us know what you think about this whole MiFast thing, and we will see you in the next one. Okay. Bye, everybody. Goodbye.