Fable 5 blocked outside the US: which Claude you can use now
Imagine trying the most powerful AI model in the world on Monday and, by Friday, watching it stop working with no warning. That is exactly what half the planet went through this week. On June 9 Anthropic opened Claude Fable 5, its most capable model so far, to the public. Three days later, on June 12, that same model vanished for almost everyone. It was not a technical glitch. It was an order from the United States government.
If you heard about Fable 5 and tried to use it from the UK, Spain, Mexico, India or anywhere outside the US, you ran into a wall. And if you are just getting started with Claude, you are probably asking a very reasonable question: "okay, so which model do I use?". Relax. In this article we explain what happened, why, who it really affects and, most importantly, which versions of Claude are still available to you and which one suits you best depending on what you want to do.
What exactly happened
The timeline is short and striking. On June 9, 2026, Anthropic launched two models at once: Claude Fable 5, designed as a safer general-purpose model with gains in coding, knowledge work, vision, memory and long context; and Claude Mythos 5, a more powerful version reserved for trusted-access users. Fable 5 came out of a testing phase that Anthropic had internally called "Project Glasswing", where it had already shown an unusual talent for finding security flaws in software.
On Friday afternoon, June 12, Anthropic received an order from the US government. The Commerce Department, leaning on national security export controls, required it to suspend all access to Fable 5 and Mythos 5 for any foreign national. Given the scope of the order, Anthropic decided to disable both models for absolutely all users. That is why they disappeared overnight.
Why the US blocked these models
Here is the part that surprises most people reading it for the first time. The reason is not commercial or financial: it is national security. According to what officials told Anthropic, the decision came after a technique surfaced for getting around Fable 5's safeguards. In the AI world, these tricks for bypassing a model's limits are called "jailbreaks".
The government's fear is that such a capable model could be used for dangerous tasks, especially in cybersecurity, if someone manages to switch off its safeguards. Remember that during testing, models in the Mythos family had already shown they could find and chain together vulnerabilities in operating systems and browsers on their own. The very power that impresses people is what set off the alarms.
Anthropic, however, does not fully share the diagnosis. The company argues that the jailbreak in question was very narrow: it would unlock certain cybersecurity capabilities of Mythos in one specific case, but it would not bring down all of Fable 5's safeguards across the board. Even so, it complied with the order.
Who the suspension really affects
This is the part that hits those of us living outside the United States hardest. The order does not only mention people in other countries. It mentions any "foreign national", wherever they are. In other words, it also affects people living in the US who do not hold American citizenship, including Anthropic's own employees.
Since splitting users into who can and who cannot was unworkable without breaking the order, Anthropic chose the safest path for itself: turning Fable 5 and Mythos 5 off for everyone. The practical result for you: if you are not a US citizen, these two models are out of reach right now, no matter which plan you pay for. It is not a problem with your account or your billing country. It is a restriction coming from above.
Anthropic's response
Anthropic has not stayed quiet. In its statement it defends an idea worth understanding: the company believes a government should indeed be able to block AI deployments it considers unsafe, but it should do so within a process that is "transparent, fair, clear and grounded in technical facts". And, in its view, this particular order does not meet those principles.
Translated for newcomers: Anthropic is not disputing that security controls over AI should exist. What it questions is the form and the technical basis of this specific decision. It is an important nuance, because it sets the tone for a debate we will hear a lot over the coming months: how far a state can go in halting a technology, and with what safeguards.
Which Claude models you can use right now
Let us get practical, which is what you came for. Fable 5 and Mythos 5 being blocked does not mean you are left without Claude. Quite the opposite. The stable lineup is intact and is more than enough for 99% of what you will do as a beginner. These are the three models available to everyone:
Claude Opus 4.8, launched on May 28, 2026, is the highest-capability model in the open lineup. It is the one you pick when you need the top level of reasoning for complex tasks. In exchange, it is the most expensive and not the fastest.
Claude Sonnet 4.6, from February 2026, is the perfect balance. Fast, affordable and, for most writing, analysis or light coding tasks, its output is practically indistinguishable from Opus. It is the workhorse.
Claude Haiku 4.5, available since October 2025, is the lightest and cheapest. It shines when you need lots of fast, low-cost responses, or when you set up tasks that repeat at scale.
Which one to choose for what you want to do
If you are starting out, keep it simple. The general recommendation is clear: use Sonnet 4.6 by default. It covers conversations, writing emails, summaries, ideas, planning your day and almost anything you can think of, without overspending and without long waits.
Step up to Opus 4.8 only when you notice that a specific task is too much for Sonnet: very long reasoning, a tangled technical problem, an analysis that demands several chained steps. It is like shifting gears on a steep climb.
And save Haiku 4.5 for the repetitive and high-volume stuff: sorting many messages, generating short responses in series, simple tasks where speed and cost come first. An easy mental shortcut: Haiku for volume, Sonnet for daily use, Opus for the hard challenges. With those three you have plenty to learn and work with Claude without missing Fable 5.
How to switch models in Claude, step by step
Knowing this puts you in control. In the Claude app, the model selector sits at the top, next to the box where you type. Click the model name and the list drops down with the options available to your account. If Fable 5 is not there, it is not an error: it is simply no longer enabled outside the US. Pick Sonnet 4.6 for day-to-day work and you are set.
A tip we always give beginners: do not switch models on every single message. Open the chat with Sonnet, work as usual and only move up to Opus when you feel a specific answer falls short. To make the comparison fair, try giving both the same task with a well-written instruction, for example: "Summarize this text in five clear points and add a two-line conclusion". You will see that in 90% of cases Sonnet gives you exactly what you need, faster and cheaper. That small habit of comparing teaches you, in practice, where each model's real limit lies, without having to take anyone's word for it, not even ours.
What this means for you as a beginner
Beyond the headline, there are two lessons worth taking away. The first is that AI has become a matter of state. When a government halts a model over national security, it is telling you this technology is no longer a toy: it carries weight in geopolitics. For you, that translates into a reassuring idea: you do not need the most powerful, most controversial model to get enormous value out of Claude. The tools you can use are excellent.
The second lesson is about dependence. This week we saw how fast a model can appear and disappear. That is why, instead of obsessing over the latest version, it pays to learn the fundamentals well: how to give clear instructions, how to structure what you ask, how to chain steps. Those skills work with Sonnet, with Opus, with Haiku and with any model that comes next. At learnaifast.io we have been hammering this point for months: what really multiplies your results is not the model, it is knowing how to use it.
If you want to turn all this jumble of names into something that genuinely saves you hours every week, that is exactly the path we teach step by step, from scratch and in plain language, in the learnaifast.io courses.
In short
Fable 5 and Mythos 5 are suspended for anyone who is not a US citizen, due to an export control order from the American government based on national security grounds. Anthropic has complied with the order, even though it disagrees with its form. And you, meanwhile, have Opus 4.8, Sonnet 4.6 and Haiku 4.5 at your disposal, which comfortably cover any need when you are starting out.
Do not let the noise of the headlines freeze you. The best time to learn how to use Claude is right now, with the tools you do have in front of you.
Want to get off to a good start without getting lost among so many names? Take a look at our Claude courses for beginners and learn to make the most of the AI you can actually use today.
