Claude Mythos 5 Returns for Critical Infra

Mythos 5 and Fable 5 differ only in their safeguards

Anthropic launched Claude Fable 5 and Claude Mythos 5 on the same day, Tuesday, June 9, 2026. They are one model underneath, and the only thing that separates them is how many safeguards are layered on top. Fable 5 was given strong safeguards to make it safer for general use.

Mythos 5 carries fewer safeguards and went only to a small group of trusted Project Glasswing partners for defensive cybersecurity work. Claude Mythos and Fable are the same model with different guardrails, released to different audiences.

Most of what followed traces back to that split. Anything aimed at the shared model hit both products, even though only one of them was open to the public.

A June 12 export-control directive took both models offline worldwide

The US government put Claude Fable 5 and Claude Mythos 5 under export controls on Friday, June 12, 2026, which required Anthropic to restrict access for foreign nationals both inside and outside the US. The rule took effect immediately, and with no way to check every user's nationality on the spot, Anthropic switched both models off for everyone.

What prompted the directive was a report about Fable 5, not Mythos 5. Amazon researchers had gotten past Fable 5's safeguards by asking it to find software vulnerabilities, and in one case it wrote code showing how one of them could be exploited. Mythos 5 went offline too because the two share an underlying model.

Anthropic's own tests later put the report in perspective. Many less capable models, Claude Opus 4.8, GPT-5.5 and Kimi K2.7 among them, found the same vulnerabilities, and every model Anthropic tested could produce the same exploit demonstration.

Mythos 5 returned first, for US critical-infrastructure organizations

The restrictions first eased two weeks after the shutdown. Washington approved a limited redeployment on June 26, and the next day Anthropic said Mythos 5, which it calls its most capable model for cybersecurity, was cleared for American groups that run and protect critical infrastructure. No one else got a suspended Claude model back that day.

The notice did not include Fable 5. On June 27 the specialist model was back for critical-infrastructure defenders while the general-use version of the same model stayed offline.

Anthropic added that it was still working with the government to open access to the wider group of domestic and international Glasswing partners. The June 27 restoration was a first step for some partners, not a return to the full launch lineup.

Fable 5 stayed offline until July 1, after controls lifted June 30

General users waited from June 12 to July 1. Anthropic announced on June 30 that the export controls on Fable 5 and Mythos 5 had been lifted. On Wednesday, July 1, Fable 5 came back for users worldwide on the Claude Platform, Claude.ai, Claude Code and Claude Cowork.

The big cloud platforms were missing from the July 1 list. Anthropic said it would turn Fable 5 back on in AWS, Google Cloud and Microsoft Foundry as fast as possible but gave no date.

Paid plans had a transition period. On Pro, Max, Team and select Enterprise plans, Fable 5 could take up to 50% of weekly usage limits through July 7, after which it switched to usage credits.

Glasswing partners got access first, with 30-day data retention

Mythos 5 did not become a public model when the export controls ended. Mythos-class access still goes through Anthropic's trusted access programs, and the official Claude Mythos page says the current model is available to vetted cyberdefenders and life scientists. A group of US organizations got it back first, and anyone outside Anthropic's trusted access programs still cannot use it.

Access also carries a condition on data. Using Claude Mythos 5.1 requires, by default, agreeing to a 30-day data retention policy so Anthropic can monitor for safety.

The timeline on the same page runs from the June 9 release through the June 12 suspension and the lifting of controls to Claude Mythos 5.1 on September 1, 2026. The model has changed since June, but who can use it has not. For an organization seeking Claude Mythos access, the trusted access programs remain the only way in.

Only Fable 5 and Mythos 5 were restricted, not other Claude models

The June 12 directive was narrow, covering only Anthropic's newest models, Claude Fable 5 and Claude Mythos 5, which were the two Anthropic suspended. Opus 4.8, Sonnet 4.6 and Haiku were left out, so everyday Claude work could shift to them while Fable 5 was unavailable.

Anthropic's own testing points the same way. Opus 4.8, Sonnet 4.6 and Claude Haiku 4.5 were among the models able to produce the same exploit demonstration the Amazon report described for Fable 5, yet none of them appeared in the directive.

What Is Claude AI? Free Plan, Models, vs ChatGPT compares the current Claude models. Anthropic's announcements tie each date to a specific model, so save a copy of the source page if you need that record.

Frequently asked questions

What is Claude Mythos 5?

Claude Mythos 5 is the version of Anthropic's newest model that comes with fewer safeguards. It runs on the same underlying model as Fable 5, and Anthropic released it on June 9, 2026 to only a small number of trusted Project Glasswing partners for defensive cybersecurity work.

Why was only Mythos 5 restored on June 27?

The US government's June 27 notice covered only Mythos 5, clearing it for a group of US organizations that run and defend critical infrastructure. Fable 5 was not in that notice, so it stayed offline until the export controls ended and returned on July 1.

Can general users access Mythos 5 or Fable 5?

Anyone can use Fable 5, which has been available worldwide since July 1, 2026 on the Claude Platform, Claude.ai, Claude Code and Claude Cowork. Mythos 5 is not open to the public, and Mythos-class access still goes only through Anthropic's trusted access programs for vetted organizations.

Are other Claude models still available?

Yes. The June 12 export controls covered Claude Fable 5 and Claude Mythos 5, the two models Anthropic suspended. Opus 4.8, Sonnet 4.6 and Haiku were outside the directive.