Skip to content
erika.taranto
ITENDE中文RU
Blog News

Claude Fable 5 is back (after they switched it off)

In just a few weeks Claude Fable 5 went from being the most powerful AI model ever released to the public to being shut down worldwide. Now it is back: what it is, why they blocked it, and what changes for you as someone who uses AI.

Erika Taranto Erika Taranto
Official selection · AI for Good 2026 11 min read
Claude Fable 5 is back (after they switched it off)
Listen to the article ~11 min

Let me start with something personal: I am glad Fable 5 is back. Claude Fable 5 is the most powerful artificial intelligence model Anthropic has ever released to the public: it launched with a lot of fanfare, a few days later it was shut down worldwide over a safety issue, and now it is available again for users on paid plans. It is a story worth understanding, because it shows how shaky the ground still is beneath the tools we use every day. I will tell it to you without the jargon, and I will add my own opinion as someone who actually works with these tools.

What Claude Fable 5 is

Let’s start with the basics. Claude is the AI assistant from Anthropic, one of the leading companies in the field alongside the people behind ChatGPT and Gemini. Every so often it releases a new model, more capable than the last. Fable 5 is one of these, and at launch it was presented as the most powerful ever put in the hands of the public.

What does “powerful” mean in practice? It means it handles very different tasks well: writing and fixing code, reasoning through complex problems, reading and understanding images, helping with research. It is not a model built for graphics or video, but it is the kind of general “brain” behind a lot of tools you may be using without even realizing it.

Alongside Fable 5, Anthropic also introduced an almost identical sibling, Mythos 5: same base model, but with some protections loosened, reserved for a small group of organizations working in cybersecurity defense. Keep it in mind, because it comes up again shortly.

The launch: off with a bang

At release, Anthropic’s message was clear: a model at the top in almost every capability test, and on top of that offered at a much lower price than its previous preview. In the AI world, where every month something comes out promising to beat everything else, an announcement like this instantly sparked excitement.

Here is my first note as someone who works with this: when a new model comes out, the excitement always runs faster than reality. “Newer” and “more powerful on paper” do not automatically mean “better for your work”. We will come back to this at the end, because it is the point that matters most.

Why so many people jumped in: websites

There is a reason why, as soon as it came out, so many people jumped on Fable 5: websites. The web filled up with people describing how they had built entire sites starting from a screenshot or a simple link, in one shot, with a result good enough to make a professional designer jealous.

Here I have to be honest, because it is a rule I always follow: I have not yet had the time to test it properly, so I am not selling you my direct experience. I am reporting what the people who used it are saying, and I take it with a pinch of salt: these are first-hour enthusiasms, not yet claims verified one by one. The fact remains that, right as the community was starting to have fun, everything came to a halt.

The shutdown: how one flaw switched everything off

A few days after launch came the cold shower. Researchers found a way to bypass the model’s protections, what in the jargon is called a “jailbreak”: getting the AI to do things it would normally refuse to do. The concern, in particular, was about cybersecurity.

A screen crossed by a digital glitch, with a red warning light on
One flaw in the protections, a jailbreak, is enough to throw the whole model into crisis.

The reaction was strong and unusual: the US government imposed a commercial-type block, and Anthropic was forced to suspend access to both Fable 5 and Mythos 5. Overnight, the “most powerful ever” model vanished from the apps.

Within a few days it went from “the best we have ever made” to “switch it off”.

Why it went down worldwide, not just in the United States

Here is the detail that caught many people off guard. The block targeted certain categories of users, but Anthropic was not able to verify in real time who each individual person behind an account was. Unable to tell them apart with certainty, it had to shut the model down for everyone: private users, companies, developers, in every country. That is how a measure born far away turned into a global shutdown. An uncomfortable but valuable reminder: the AI tools we use are not ours, they live on platforms that can change the rules, or turn off the taps, without warning.

The return: Fable 5 is available again worldwide

After weeks of back and forth, and daily articles asking “is it coming back or not”, the situation cleared up. The block was lifted, and Anthropic announced that Fable 5 is available again worldwide for users on paid plans, with no more need to verify the nationality of whoever uses it. The company put the model back in service with a targeted protection against that flaw; for some routine tasks, like writing code, certain requests are for now handled by the previous model while the filters are being refined.

A studio desk at dawn: the power button lights up pink and the screens wake back up
The return, with a targeted protection against that flaw.

A practical note: “available worldwide” is the announcement, but the actual rollout arrives in stages. So Fable 5 may not have shown up in your model menu yet, or you might see it greyed out with “not available”: it is not your mistake, and it is not a door closed in your country, it just has to reach you too. Usually it is a matter of a short wait.

My personal impression, and it stays just my idea. I am happy it is back, but in my view they are bringing it back a bit more “controlled” than it was at launch. There is one concrete clue: it returned with a targeted protection against that flaw and, for now, some routine requests are handled by the previous model while they refine the filters. My feeling is that, if you use it to push yourself into certain kinds of research, you will find it more cautious than before. I do not have the proof in hand, it is a guess of mine, we will check it by using it. And in the meantime, I admit it, I cannot wait to get my hands on it.

What changes for you as someone who uses AI

Let’s land all of this on what is actually in it for you.

  • If you have a paid plan of Claude, in the coming days you might see Fable 5 appear among the selectable models. If it is not there yet, it is only a matter of waiting.
  • Wherever you are the return applies the same way: there is no block tied to your country. If it does not show up right away, it is just the gradual rollout, give it a little time.
  • If you use a free plan, for now you stay on the standard models, which are perfectly fine for most everyday work.

The exact conditions (which plans, which limits, in which countries) keep changing during these weeks, so the right reference is always the official Anthropic page, not the sensationalist headlines.

What I take away from this story

Beyond the single model, this episode says two things I always keep in mind in my work.

The first: we use the tools, we do not own them. A model can disappear for weeks because of decisions that are not up to us. That is why I never tie a project to a single tool, and I always keep my materials safe, independent of whatever platform is in fashion.

The second: the newest model is not necessarily your model. Fable 5 is powerful, but “powerful” and “right for what you need to do” are two different things. Before turning your way of working upside down to chase the latest release, you test it on your real case and judge by the results, not by the hype. It is exactly the method I use to choose tools when I work for a client: not the most talked about, the most suitable.

Sources

Did you like it? Share it.
Comments

Leave a comment

Comments are reviewed and approved before they appear.

Your email will not be published.

FAQ

Frequently asked questions

Does Iris really exist? +
No, Iris is a character. But she is not invented out of nothing: she is inspired by real blind painters like John Bramblitt and Esref Armagan, who paint by touch. And behind her are 285 million people in the world who live with a visual impairment.
Did the AI make the film instead of you? +
No. The AI is the visual tool, not the director. I wrote the story, directed every shot, chose the movements, the voices and the edit, and discarded most of what I generated. The tool executes, the decisions are mine.
Which tools did you use? +
Higgsfield Cinema Studio and Kling for images and movement, ElevenLabs for the two voices, CapCut for editing and music (library, royalty free, not AI generated).
Why does the AI voice sound so robotic? +
It is intentional. I could have made it warm and human, I chose not to. The film is an honest dialogue between a person and a machine: I wanted you to feel that it is a machine, without pretending otherwise.
Does an AI that describes colors really exist, or is it science fiction? +
It already exists today, and it is free: Be My Eyes, Seeing AI, Envision. The film only imagines how this technology could enter an artist's studio. It is not science fiction, it is the present.
Where can I see UNSEEN? +
UNSEEN is screened at the AI for Good Film Festival in Geneva on July 9, 2026, where it is among the ten finalists. After the event I will make it available online: the best way not to miss it is to follow me on my channels.
How much did it cost to produce? +
Much less than you would imagine for a short: I made it alone, without a crew and without a major studio budget. And that is exactly the point: today a quality of storytelling that required months and out-of-reach figures is within reach of a single person.
Are the voices cloned from real people? +
No. They are synthetic voices, not clones of existing people. The music is from a library, not AI generated.
Can I use AI like this for my brand too? +
Yes, and that is what I do. The same approach that took UNSEEN to Geneva I use for ads, images and AI faces for brands. If you want to talk about it, you can find me in the contact section.