Anthropic announced that Claude Fable 5 will be redeployed globally starting July 1 after U.S. export controls imposed June 12 shut down access to Fable 5 and Mythos 5 for everyone due to immediate compliance needs. The controls were lifted June 30, with Fable 5 returning across Claude Platform, Claude AI, Claude Code, and Claude Cowork for Pro, Max, Team, and select Enterprise plans, included for up to 50% of weekly usage limits before shifting to paid usage credits after July 7. Mythos 5 access is being restored only to certain U.S. organizations via Project Glasswing following June 26 approval. The shutdown followed a report that Amazon researchers found a method to bypass Fable 5 safeguards; Anthropic says the bypass is now blocked in over 99% of cases and that U.S. Commerce testing agreed the new safeguards are extraordinarily strong, alongside plans for deeper government collaboration and pre-release evaluations.
00:00 Fable 5 Returns
00:45 Global Rollout Details
01:13 Usage Limits and Credits
01:54 Mythos 5 Partial Restore
02:32 Why It Was Shut Down
03:46 Timeline and Big Picture
06:19 Amazon Bypass Explained
08:06 New Safeguards Breakdown
09:11 Industry and Government Frameworks
10:05 Wrap Up and Community Plug
Full transcript
We've got some fantastic news today for AI lovers as Fable 5 will be redeployed by Anthropic and this is the official announcement this whole news article just dropped like a few minutes ago so essentially what has happened here is on Friday June the 12th US government shut down called Fable 5 and Miphos 5 export direct control goes on that and nobody can use it that was fully taken down as of today June the 30th the export controls on Fable 5 and Miphos 5 have been lifted and Fable 5 the headline is Fable 5 is coming back tomorrow so within the next sort of 12 to 16 hours I would expect to see Fable 5 access coming out again if you're wondering okay will it be available for everyone globally yes it will that's another good piece of news right here so we can see the news here that Fable 5 will be available starting tomorrow aka first of June sorry first of July Wednesday to users globally so around the whole world not just the US on Claude platform Claude AI Claude code Claude co-work for pro max team and select enterprise plants now Fable 5 will also be included for up to 50% of weekly usage limits what that means essentially is the up to 50% of your weekly usage on your Claude subscription can be used with Fable 5 which means it's included inside the subscription however what they've actually said is after that it will be available via usage credits which is basically like I think you have to pay to get access to it um to continue using it right you have to pay for usage credits and then you have to reload your balance every time you run low so after the after July the 7th then you'll have to pay for it which is the bad news but we already knew that was coming eventually anyway so what they've also said is they've restored access to Miphos 5 for a set of US organizations not really relevant to us but basically what they're saying here is that following the US government's approval on June the 26th this will be rolled out again to Project Glasswing which is a select number of cybersecurity firms and enterprise companies who will have access to Claude Miphos 5 now also they've gone on like a long tangent explaining everything that happened and I'll cover that in a second I've not actually read through that so we're reading it through together because this is brand brand brand new but what we have over here is the full backstory on this so if you're fully new to this just to recap for those at the back and what just happened on Friday June the 12th the US government applied export controls to two of their models so Claude Fable 5 and Claude Miphos 5 and this required Anthropic to basically restrict access to foreign nationals whether inside or outside the United States now because the order took effect immediately and they had no reliable way to really check if someone is in the US or outside the US or they're from from the US or not from the US they were just like right let's shut it down for everyone now that has been lifted right so you can see the announcement over here from Howard Lutnick who says over the past two weeks we have worked closely with Anthropic to analyze and approve Fable 5 to ensure alignment across the US government and strengthen America's leadership in AI now this is good news Susie Wiles also announced it as well as you can see here so under President Trump's leadership in the United States apparently they are the undisputed winner in the AI race so I take that for what you will but what I'm saying here is basically the the export control got lifted then Anthropic made an announcement and that's where this all came from so let's talk about the timelines now and what happened so they've actually provided further details and updates in four different areas so number one a timeline of events including updates have made to the safeguards so they've actually added some updates to the safeguards on top of that they've added a new section and some updates on general approaches to safeguards then they've talked about a shared industry framework so what they actually said here is that because of the whole shutdown but by the way the shutdown was triggered by someone announcing a jailbreak of Fable 5 now there's no confirmation of what happened from that from what I can see so far but basically there were rumors that a Chinese company got access to it there were other rumors the Amazon's team actually managed to jailbreak it they reported it and that's why it got shut down nothing's been fully announced on that so maybe they've talked about that inside the announcement here we'll take a look at that in a second and they also said together with Amazon, Microsoft, Google and other Glasswing partners we've started to develop a framework to fix these potential problems where people bypass the safeguards of these models and then they've also talked about deeper government collaboration so I think one thing we're going to see is number one in the future this is not the first time it's going to happen it's not the first time a frontier model is going to get removed number two this is good news because we were all worried about gated stuff but number three the US government especially is getting really involved in these AI models and I think we're going to see that progress as a theme over the next few releases because we're at a point now where AI is so powerful and so much better than it was a couple of years ago that the US government is like going to get really cracking down on like okay who this gets released to who has access how many safeguards or what safeguards are on the models as well so I think that's the theme we'll see in the future so if you look for example with OpenAI and GPT 5.6 that is not available to anyone right it's only available to like 20 different enterprises nobody like me has access to it you know and it's one of those where we're going to see that more and more why is it is because the government is like really cracking down on who gets access which is kind of bad for Anthropic and for OpenAI because if they're trying to compete in a world you know globally where China is just open sourcing all their projects and just releases whatever they want whenever they want it's going to be pretty hard for them to keep up but we'll see that's a totally different topic for a totally different video so we have the timeline and safeguard effects so June the 9th got released June the 12th got shut down and the reason are so this is actually because of Amazon yeah so I said I haven't read this report but yeah it is down to Amazon so that's why it got shut down so the export control directive on June the 12th came after the government became aware of a report in which Amazon researchers had found a method of bypassing Fable 5 safeguards prompting it to find vulnerabilities and so that's why it got shut down okay interesting so basically Amazon cracked it they found a way to bypass Fable 5 and that's why Fable 5 is no longer accessible to any of us until tomorrow and then they've also said the with many less capable models like GPT 4.8, 5.5, Kim K 2.7 they can be bypassed in the same way but the reported technique for the bypass didn't expose any unique MIPHOS level cyber capabilities so it wasn't like MIPHOS level stuff it was just like basics or bypasses so they started working with the government to fix it and the classifier that Amazon found when they bypass the model is now blocked in over 99% of cases in a very small fraction of cases the model may provide information that isn't detailed enough to help someone anyway but they've talked about their setup right here and also researchers from the US department of commerce's center for AI standards and innovation have tested both prior new safeguards and agree that they are extraordinarily strong apparently so then they talked about their approach to safeguards blah blah blah some people as well might react to this and say is this a good thing that Fable 5 is released to everyone so I think they've really taken safeguards seriously Fable 5 isn't as strong or as powerful as MIPHOS 5 it's not focused on cyber security in the same way the main theme of this seems to be classifiers so that seems to be the bypass method that some of these instances came from and sometimes they can fail but they've set up a bunch of safety classifiers to trigger if someone is trying to use this technique which is pretty crazy so basically I think they've got like early warning signals in case anyone was trying to use it for something dodgy and then they've broken it down in terms of the safeguards here so they've got classifier boundaries in terms of what's allowed and what's blocked basically what they're trying to do here is create a larger margin of safety before anything is bypassed so the normal safeguards are very different to Fable 5 safeguards then they've talked about like a consensus for industry frameworks it seems like they're just working with other project glasswing partners to improve like a better way to avoid these issues which is going to become like something pretty serious in the future especially as the serious in the future especially as these models get powerful bear in mind like Fable 5 is the worst it's ever going to be right if Miphos is the worst it's ever going to be it only gets more powerful from here and then also you can see what they're going to do in the future so since they've started working with the US government closely what they've said is that everything will go through pre-release so this is the new normal now it's like everything will be gated everything will go with pre-release betas and then once it's evaluated by the government then it will get released and then there'll be rapid information sharing on safeguards dedicated resources for joint research and a common industry bar for safeguards as well across the industry so that's basically it for the whole announcement the good news is Fable 5 is coming back it's available globally doesn't seem like you need identity verification to use it which is something that people worried about as well whatever happens we just make better systems each day and that's the thing that i would recommend for you is like it doesn't matter what happens next doesn't matter which models get released which ones don't what you want to focus on is are your systems good enough so what we've actually done is we've built the agent operating system with all of our best systems as you can see so we've got a video agent an seo agent everything that i work on day to day and all my agents are plugged into one system with a shared memory so that we can build and automate anything if you want to get the same system in the mission control dashboard you can get that inside the AI profit boarding link in the comments description or go to the air profitable.com this is my community for learning and growing and scaling with AI we have loads of tutorials already inside here on Fable 5 and as soon as Fable 5 gets released we'll plug it into our agent operating system as well if you want to get access to that link in the comments description you can also ask questions inside the community and i personally answer them with a video tutorial every single day on top of that inside the classroom you get access to all my best trainings and we have a new daily update section here where as soon as we learn something new and useful we share it as a video tutorial with a step-by-step guide and we date it so you can see what the latest stuff is and you can also grab the agent operating system i just showed you with a video tutorial the last update day a step-by-step guide and the zip file to install over here if you want weekly automation coaching calls on how to automate your business to save time with AI we have four weekly coaching calls inside here you can jump on the calls share your screen see what other members are working on connect with other people and also inside the map you can meet people locally who are building with AI agents like you so feel free to get that link in the comments description or just go to the aiprofitborm.com thanks for watching
More episodes