Why Claude Fable 5 is not working: Anthropic suspends it after US government order

News · 13 Jun 2026 · Claire Bennett

If you have opened Claude over the past few hours and found that Claude Fable 5 has vanished from the model picker, or your app is throwing an error when you try to use it, you are not imagining it. Anthropic disabled both Claude Fable 5 and Claude Mythos 5 for every customer worldwide on the evening of 12 June 2026, after the US government issued an export-control directive citing national security. The models are unavailable right now, and Anthropic says it is working to bring them back.

The short version for anyone searching “why is Claude Fable 5 not working”: it has not been deleted, and you have not lost your subscription. A government order forced Anthropic to switch off these two specific models. Every other Claude model, including Claude Opus 4.8, still works normally. Here is what happened, who it affects in the UK, and what to do in the meantime.

Why Claude Fable 5 is unavailable right now

On 12 June 2026, Anthropic published a statement saying it had received a US government directive at 5:21pm Eastern Time that day. The directive, which Anthropic describes as an export-control order issued on national security grounds, ordered the company to suspend all access to Fable 5 and Mythos 5 for any foreign national, whether inside or outside the United States, including Anthropic’s own foreign-national employees.

Because Anthropic could not reliably separate foreign nationals from everyone else on a per-user basis, it took the only practical step available and disabled both models for all customers globally. The company’s own status page logged the change at 00:50 UTC on 13 June, marked as an active incident. Anthropic apologised for the disruption, said it believes the order is a misunderstanding, and promised to “share more details over the next 24 hours”. As of the time of writing, no further detail or restoration has been published.

What the government says, and what Anthropic disputes

This is where the two sides diverge, and it is worth reporting both plainly. According to Anthropic, the government’s stated basis is that it had become aware of a method of “jailbreaking” Fable 5, meaning a way of bypassing the model’s safety controls. The US government has not published its own technical account, so the only detailed description available is Anthropic’s. That asymmetry matters: most of what is publicly known about the reasoning comes from the company that was ordered to comply, not from the authority that issued the order.

Anthropic says it reviewed a demonstration of the technique and concluded it surfaced a small number of previously known, minor vulnerabilities rather than a universal bypass. The company characterises the method narrowly, describing it as “asking the model to read a specific codebase and fix any software flaws”, and argues the same capability is “widely available from other models”, naming OpenAI’s GPT-5.5 directly. Anthropic also says that across thousands of hours of red-teaming with the US government, the UK’s AI Security Institute (AISI) and outside organisations, no universal jailbreak was found, though it notes the UK AISI “made progress towards one”.

The company’s broader objection is a policy one. It warns that if a single demonstrated vulnerability were enough to pull a frontier model, “it would essentially halt all new model deployments for all frontier model providers”. Anthropic also used the statement to defend its 30-day data-retention policy for these models, which it says carries real cost but lets it research and mitigate jailbreaks. None of this changes the legal position: Anthropic has complied with the order while publicly disagreeing with it. Major outlets including CNBC, NBC News and 9to5Mac reported the suspension on 12 June.

Who is affected in the UK, and what still works

The suspension is global, so UK users are caught by it in full. If you pay for Claude.ai on a Pro, Max, Team or Enterprise plan, Fable 5 has disappeared from your available models. Developers calling Fable 5 or Mythos 5 through the Claude API, Amazon Bedrock, Google Vertex AI, Microsoft Foundry or GitHub Copilot will see the model return errors rather than responses, and any integration hard-coded to Fable 5 needs updating.

The important reassurance is that everything else is untouched. Anthropic has been explicit that access to all its other models is unaffected, and new sessions fall back to Claude Opus 4.8 or whichever model you had selected before. So Claude itself is not down. Your projects, chat history, tools and settings all remain. Only the two newest Mythos-class models are switched off. If you only ever used the default Claude model, you may not notice any change at all.

What UK users and businesses should do now

The simplest move is to stay inside Claude and switch to Opus 4.8, which Anthropic itself points users towards. It is a highly capable model and the officially supported fallback, so most everyday writing, analysis and coding work will carry on with little disruption. If you run automated workflows or agents that name Fable 5 explicitly, update those model IDs now to avoid silent failures, and check any Bedrock, Vertex, Foundry or Copilot configuration that referenced it.

If your work genuinely depended on the extra capability Fable 5 offered, it is reasonable to trial an alternative in parallel rather than wait. OpenAI’s ChatGPT, running GPT-5.5, and Google’s Gemini are the obvious comparisons, each with different strengths in coding, research and ecosystem integration. We have compared the main options for UK readers in our guide to Claude vs Copilot vs Gemini, and broken down the plans in our ChatGPT UK pricing guide. None is a like-for-like replacement for Fable 5’s specific tuning, so the sensible step is a short side-by-side test with your own prompts before committing.

On timing, Anthropic has said only that it is “working to restore access as soon as possible” and that it would share more within 24 hours of its statement. Treat that as the company’s stated intention rather than a firm date, because the decision ultimately rests with the US government, not with Anthropic. Anyone weighing up whether to keep paying for a plan they took out partly for these models should know that the rest of the Claude service continues as normal in the meantime. For the current cost of each tier in pounds, see our Claude UK pricing breakdown.

What Fable 5, Mythos 5 and Project Glasswing actually are

If the names mean nothing to you, the timing explains why: both models are only days old. Anthropic launched them on 9 June 2026, three days before they were pulled. Fable 5 was billed as Anthropic’s most capable widely released model, priced at $10 per million input tokens and $50 per million output tokens, which the company said was less than half the price of its earlier Mythos Preview.

Mythos 5 is the same underlying model as Fable 5, but with some safeguards lifted in certain areas. It was deployed through what Anthropic calls Project Glasswing, a collaboration with the US government, and was already restricted to vetted partners rather than the general public. Fable 5 was the safer, public-facing version: in three sensitive areas, cyber, biology and chemistry, and model distillation, it automatically routes flagged requests to the older Claude Opus 4.8 instead. Anthropic says more than 95% of Fable sessions involved no such fallback at all. The episode lands against a backdrop of growing scrutiny of frontier AI on both sides of the Atlantic, from Anthropic’s own ASL-3 safety protections to the UK’s push for sovereign AI capacity. For now, the practical reality is simpler: two new models are off, the rest of Claude is on, and the next official word is expected from Anthropic shortly.