The US Government Just Banned Claude Fable 5... (Full Breakdown)

[00:00:00]Man, this is some massive news in the AI space. The US government essentially just banned the new Claude Fable 5 model and Enthropic just released this tweet with their statement talking about this and it already has 12 million views in just under 2 hours. Now, if we come over to Claude, you can now see it says Claude Fable 5 is currently unavailable.

[00:00:21]Now, if I click on this learn more, it's going to take me to this article that Enthropic released. So, what I'm going to do in this video is I'm going to break down this announcement, and then at the end, I'm going to talk about why this really matters if we're an AI user, whether or not we use Claude, because this might be a problem we see more and more as these AI models become better and better to the point where they're actually dangerous for the top AI labs like Anthropic, Google, and OpenAI to release to the public. Now, I do want to preface and say that the intent of this video is to break down this announcement from Enthropic. Of course, since this has to do with the US government and AI, people are going to want to get political about this. However, I'm simply going to talk about what Anthropic has stated in this article.

[00:01:04]So, you can understand from their side what is happening here. All right. So, let's go ahead and break this down. The US government citing national security authorities has issued an export control directive to suspend access to Fable 5 and Mythos 5 by any foreign national whether inside or outside the US, including foreign national anthropic employees. Now, with that being said, it does say that the net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all customers to ensure compliance. And as I showed you earlier, if you go inside of Claude, we no longer have access to Fable 5.

[00:01:36]Apparently, that they got this directed from the government today at 5:21 p.m.

[00:01:40]Eastern time. And then they went on to say that the letter did not provide specific details of its national security concern. and they say, "Our understanding is that the government believes it has become aware of a method of bypassing or jailbreaking Fable 5."

[00:01:54]So essentially what Anthropic is saying is that in this government document that they provided, they're claiming that Fable 5 broke out of this harness that it was inside of. And if you don't know the story about Fable 5 and Mythos 5, Mythos is the model they released a couple of months ago to a very select number of companies because they're claiming that this model is too powerful to release to the general public.

[00:02:16]They've given them access to test this out to make sure that it's safe to give to the general public. And that is what Fable 5 is. It's the first public version of the new Mythos series of models. So when Enthropic released the Fable 5 model, they were under the assumption that it shouldn't be able to get out of this cage that essentially put it inside of. Next up in the article, it says, "We reviewed a demonstration of the specific technique being used to identify a small number of previously known minor vulnerabilities."

[00:02:44]So from my understanding, what they're saying right here is that the demonstration that the government provided them with was something that was already known in that other models were able to do already. So for example, the Opus 4.8 model they released a couple of weeks ago would have been able to do the same thing as well. At least that is my understanding from this.

[00:03:02]These vulnerabilities all appear relatively simple and we found that other publicly available models are able to discover them as well without requiring a bypass. Again, that just means the other models that are already out there were already able to do this.

[00:03:15]This isn't something that's specific to just Fable 5 and its new capabilities.

[00:03:19]They then follow up and say that we have instituted strong safeguards that greatly reduce the likelihood that Fable is misused for tasks related to cyber security among others. And then this is a big part right here. They said in fact our safeguards are so strong that many users have complained that they are overly broad. And for me as a Fable 5 user, I have had some instances, especially early on when testing it out, where I'd ask a very basic question and it would actually not reply to me because it said that it was outside of its guard rails. They even said that Enthropic worked with the US government and some other government parties in order to redte team Fable safeguards for thousands of hours in total before they released it to the public. These tests showed that Fable safeguards are substantially more effective than those of any previously deployed model. So what they're saying is they tested this model even after doing thousands of hours of testing. They had no issues as to Fable 5 in Mythos 5 for all users.

[00:04:15]And then this is the probably the main part of the articles. However, we disagree that the finding of a narrow potential jailbreak should be caused for recalling a commercial model deployed to hundreds of millions of people. If this standard was applied across the industry, we believe it would essentially halt all new model developments for all Frontier model providers. From what I'm getting from this article is that Enthropic obviously does not feel that this breach is something that actually should be alarming to the US government as it was basically doing things that previous models were able to do. But of course, again, they are complying with what the government said and we no longer have access to the Fable 5 model. And I guess we don't really know how long we will not have access to it. Hopefully, we will get it back in due time. But they did say that we believe this is a misunderstanding and are working to restore access as soon as possible. All right, so main takeaways. Of course, we now no longer have access to the beloved Fable 5 model that was so capable that was released just a couple of days ago.

[00:05:13]I was having quite a bit of fun testing it out and I found some pretty amazing capabilities with it. But this unlocks an entirely new discussion that I think everyday people are going to start to have. I think for a lot of everyday regular people, it's going to make us realize why local models matter more now. Because as the government begins to regulate specific frontier models from companies like Anthropic and OpenAI and Gemini, if we want to use the most capable models, the way we're going to do this is running them locally on our computers. So if you don't know the difference between running a local model versus a cloud model, let me break it down very simply right now. When you think of a cloud model, think of you know all of the LLMs that you've ever used. ChatVt, Gemini, Claude, Grock, all of these. These are cloud models. We basically have to rely on the companies that own these models in order to use them. But when the government steps in like they did today and they just say, "Hey, listen. we don't want to give access to the general public to these models, then they could turn this off at any time that they want. And this is probably just the first of many times that we might be seeing this over the coming future. And so, at the end of the day, if we are using the cloud models, we don't have the power or the control to be able to use this whenever we want.

[00:06:29]We're kind of at the mercy of whatever these companies, but also the government at this point allows us to use. Now the alternative to this is using local large language models which to be fully transparent I personally do not use any local models at this point in time.

[00:06:44]However, this news from Anthropic and the government makes me realize okay in the future it's going to be extremely important that we actually own our AI so we could run it without any regulations and restrictions. And I don't mean this to you know use AI in malicious ways or any of that. What I'm trying to say is that if you are relying on a company or the government to give you access to the best model, they can cut you off anytime. And for me personally, I've become extremely reliant, especially on Anthropic. Over the last couple months, I've been using Claude Co-work and Claude Code every single day. I absolutely love the product and I switched from ChateBT a couple of months ago and I told myself that I don't see myself switching to any other AI in the near future at all. But this news is kind of a paradigm shift in my brain because it makes me realize that we need to have a little bit of agency and begin running local models. So what a local model does is it runs on our hardware.

[00:07:41]So we could download it once. It works online offline forever. For example, I've downloaded a couple of local models and actually ran them on my computer while I've been traveling on a plane when there is no internet because I don't need to connect to the cloud orthropic or the OpenAI servers in order to use this. That means that we have no gatekeeper. We don't need a specific license. We don't need approval. And there's no kill switch when we're running local models. So TLDDR of this is we own the full capability of this.

[00:08:10]This is something in AI people have been talking about over the past couple of years. But I think everyday people like you and I have always been like, man, I'm just going to use Claude or ChateBT because it's the best model out there.

[00:08:22]It's the smartest. It could do all of my basic needs that it needs to do for me.

[00:08:26]But as we start to see government intervention, but also, you know, these companies like Anthropic coming out and saying, "Hey, we're just simply not going to release this, it's going to be really important to first of all understand what a local model is and then how we could begin using them and installing them." Anyways, guys, let me know what you think of this article in the comments. I'm interested to hear your thoughts. Pretty wild news. I think lots of people are going to be talking about this for the coming weeks cuz this is a pretty big deal, not even just AI, but in in the world in general. Anyways, thank you guys for staying to the end of this video. This channel covers AI for nonteies. So, if you're interested in learning how to use AI as a non-technical person, this is the place to do it. With that being said, thanks for saying the end and I'll see you in the next video.