Update, 21 July 2026: The free Fable 5 window has closed. It ended on 19 July 2026 at 11:59:59pm PT and there was no fourth extension. From 20 July 2026, Fable 5 is a standard inclusion on Max plans and on premium Team and premium legacy Enterprise seats, and runs on pay-as-you-go usage credits on Pro plans and standard seats. Full detail in the status box below. Background in our earlier report, Claude Fable 5 stays free for paid users until 19 July.
Claude Fable 5 is working, globally, and has been since 1 July 2026. What changed on 20 July 2026 is how you pay for it. The free promotional window closed on 19 July 2026 at 11:59:59pm PT and was not extended again.
On Max plans, on premium seats on Team plans and on premium seats on legacy seat-based Enterprise plans, Fable 5 is now a standard part of the plan, usable for up to 50 per cent of weekly usage limits at no extra cost.
That 50 per cent is carved out of the same weekly pool every other model draws from, so it is a ceiling rather than extra headroom, and Anthropic states that Fable 5 draws down usage faster than its other models. On Pro plans, on standard Team seats and on standard legacy Enterprise seats, Fable 5 runs on pay-as-you-go usage credits, with a one-time usage credit to start you off.
Claude Code needs version 2.1.170 or later to offer Fable 5. Mythos 5 is a separate case: it has only been restored for a set of US organisations. We update this page as the situation changes.
Update, 1 July 2026: Fable 5 is being redeployed worldwide
Anthropic has confirmed that Claude Fable 5 is coming back. In its post Redeploying Claude Fable 5 (30 June 2026), the company states: “As of today, June 30, the export controls on Fable 5 and Mythos 5 have been lifted.”
On who gets access and when, Anthropic says: “Fable 5 will be available starting tomorrow, Wednesday, July 1, to users globally on the Claude Platform, Claude.ai, Claude Code, and Claude Cowork.” That means Australia is included this time, not left waiting.
On the usage side: “For Pro, Max, Team, and select Enterprise plans, Fable 5 will be included for up to 50% of weekly usage limits through July 7, after which it will be available via usage credits.”
[Update, 21 July 2026: Anthropic later extended that window to 19 July 2026, when it closed for good. From 20 July, Fable 5 is included as standard on Max plans and on premium Team and premium legacy Enterprise seats, for up to 50 per cent of weekly usage limits, while Pro plans and standard seats moved to pay-as-you-go usage credits.]
For the big clouds Anthropic adds: “We will re-enable access on AWS, Google Cloud, and Microsoft Foundry as quickly as possible.”
Mythos 5 is more limited. Anthropic says: “We have also restored access to Mythos 5 for a set of US organizations, following the US government’s approval on June 26.” So Fable 5 is back for everyone worldwide, while Mythos 5 stays restricted to select US organisations for now. The rest of this page kept below documents the original 12 June suspension and the fixes people searched for while it was down.
Archive: the June 2026 suspension
What follows documents the June 2026 suspension: the error people hit, the fix they ran, what the US directive actually said, and what the episode still tells Australian teams about depending on one vendor. It is a record of that period, not a live outage report. For the current position, see the status box at the top of this page.
If you opened Claude Code in mid-June 2026 and found yourself booted off Claude Fable 5, your laptop was fine and your account was not broken. The cause sat a long way from your terminal. Anthropic had suspended two models for every customer, worldwide, to comply with a US government directive.
The exact message many developers were staring at read: “There’s an issue with the selected model (claude-fable-5). It may not exist or you may not have access to it. Run /model to pick a different model.” That string was being pasted into Google from the United States to Australia. It was not a setup problem, and there was a one-line fix.
The one-line fix people ran during the outage
The model your session pointed at had been pulled, not your install. There was no need to reinstall, log out, rotate API keys, or touch your network. All that changed was which model answered you.
In Claude Code, run /model and switch to Claude Opus 4.8, Claude Sonnet 4.6, or whatever your previous default was. Those models remain fully available, so your session picks straight back up. If the picker misbehaves, run claude update first, then try /model again. Projects, history and usage limits are untouched.
Which models are affected
It is worth naming both so nobody assumes this is a Fable-only event. The affected models are Claude Fable 5 and Claude Mythos 5. Every other Anthropic model keeps working, including Opus 4.8, which is exactly why the workaround above is reliable.
New sessions across Claude products fall back to your selected default, or to Opus 4.8 if you have none set, so most products degrade gracefully rather than failing hard. At launch the two were not equal: coverage noted Mythos 5 was limited to a small participant group while Fable 5 was the public version.
What actually happened
The trigger was not a bug or an outage. According to Anthropic’s published statement, it received a US government directive on 12 June 2026 at 5:21pm ET and moved to comply.
The scope is the key point, and it is global. The directive covers “any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.” It is not aimed at Australia, at Spain, or at any single country. Anthropic says it cannot filter foreign nationals from everyone else in real time, so it disabled both models for all customers to stay compliant. That is why even US citizens could not use them at the time.
On who sent the order, Anthropic’s own statement says only “the US government.” Press reporting goes further. The directive came as a letter from Commerce Secretary Howard Lutnick to Anthropic chief executive Dario Amodei, drafted with the Commerce Department’s Bureau of Industry and Security, as reported by Axios and corroborated by Bloomberg and CNBC. Those names come from the outlets, not from Anthropic, and no outlet has named a specific statute by number.
The dispute over a jailbreak claim
This is where the two sides diverge, and it is unresolved. The US government’s stated reason is that it believes it became aware of a method of bypassing, or jailbreaking, the model. That reason has not been independently verified, and nobody should treat any bypass as settled fact.
Anthropic pushes back directly. Its position is that the demonstration disclosed only “a small number of previously known, minor vulnerabilities,” and it adds that other publicly available models can discover the same issues without a bypass. The most that has been reported about the disputed task, attributed to Anthropic via the press, is asking the model to read a specific codebase and fix software flaws. The honest framing is a live dispute, with both sides on record and neither verified.

Why Australian businesses should care
Here is the part that should make Australian boards sit up. Australia is a member of AUKUS and Five Eyes and one of Washington’s closest allies, yet Australian developers and Australian-based staff are still “foreign nationals” under a worldwide order. This is not the US cutting off Australia. It is a global measure that catches a close ally with no exemption, the same way it catches every other foreign national.
AI adoption among Australian businesses has climbed sharply, and the uncomfortable lesson is that a model your team built on can vanish on a foreign government’s order overnight. The response is continuity, not panic. Configure a multi-vendor fallback, set a default that is not a brand-new release, and review vendor agreements for kill-switch and model-deprecation clauses.
For teams handling sensitive records, this also sharpens the data sovereignty question that already sits under the Privacy Act 1988 and OAIC guidance: who controls the model, and under whose laws. Building resilient conversational AI and document automation means treating model availability as a supply-chain dependency, like any critical SaaS vendor.
Was it temporary, and when did Fable 5 return
Anthropic called the situation a misunderstanding and said it was “working to restore access as soon as possible.” The suspension was presented as temporary, and nothing in any source said these models were gone for good. At the time, no firm return date existed.
One administration official told the press the lockdown should hold until US national-security systems were “hardened,” which the official suggested could happen in the coming weeks. That language was soft and non-binding, so no date was guaranteed then. The answer turned out to be 30 June 2026, when the export controls were lifted, with Fable 5 back for users globally the following day.
Where things stood on 24 June 2026: the models were still offline for everyone. An Anthropic executive had said on 21 June that access could return “in the coming days,” but that had not happened by 24 June.
The clearest marker on the calendar then looked like 8 July, when Anthropic’s identity verification was due to take effect: that could have let Fable 5 return for verified US users first, while Australia and other international users stayed on Opus 4.8. A broader fix that included allies looked more likely to hinge on a US government framework due around 1 August, and prediction markets were leaning towards a restart in the middle of July.
None of that is how it resolved. The controls were lifted on 30 June 2026, ahead of both markers, and Fable 5 came back on 1 July 2026 for users globally with no nationality check.
Is Claude Fable 5 down for everyone?
No, not any more. It was disabled worldwide for all customers on 12 June 2026, but Anthropic lifted the controls on 30 June and restored Fable 5 for users globally from 1 July 2026. It has been available ever since.
What ended on 19 July 2026 was the free access window, not access itself: from 20 July, Fable 5 is a standard inclusion on Max plans and on premium Team and premium legacy Enterprise seats at up to 50 per cent of weekly limits, and runs on usage credits on Pro plans and standard seats. Mythos 5 remains limited to a set of US organisations.
Which model should I use instead of Fable 5?
Fable 5 is available again, so you may not need a substitute at all. If you do want one, or if you are on a plan where Fable 5 now draws on usage credits, run /model in Claude Code and switch to Opus 4.8 or Sonnet 4.6. Both remain fully available and your work carries on unaffected.
Is Fable 5 back yet?
Yes. Anthropic says the US export controls were lifted on 30 June 2026, and Claude Fable 5 became available again from Wednesday 1 July 2026 to users globally, including in Australia, on the Claude Platform, Claude.ai, Claude Code and Claude Cowork. It has stayed available since.
Since 20 July 2026 it is a standard part of Max plans, premium Team seats and premium seats on legacy seat-based Enterprise plans, for up to 50 per cent of weekly usage limits at no extra cost, while Pro plans and standard Team and legacy Enterprise seats use pay-as-you-go usage credits, with a one-time usage credit to start. Claude Code needs version 2.1.170 or later to offer it. Mythos 5 has only been restored for a set of US organisations.
When will Fable 5 come back?
It already has. Anthropic redeployed Fable 5 to users globally from 1 July 2026, after the US export controls were lifted on 30 June 2026.
Did Australia get singled out?
No. The directive applies to all foreign nationals worldwide. Australia is one example of who is caught, not a specific target.
What this means for AU teams
If a single directive can pull a model your team was building on, the question for Australian operators is not which model to trust, but how many you can switch between without losing a day. That is a continuity decision worth making before the next outage, not during it. Designing that multi-vendor setup, with a sensible fallback and a clear AI visibility strategy, keeps one provider’s bad week the provider’s problem rather than yours.
