Fable 5 is back, but the government left a heavier safety brake on it

Anthropic is bringing Claude Fable 5 back after a US government intervention. The company says export controls imposed on June 12 were lifted on June 30, and Fable 5 is available globally from July 1 across Claude Platform, Claude.ai, Claude Code and Claude Cowork.

Fable 5 returned after an eighteen day government stop

The source item is Zvi Mowshowitz's commentary, but the core facts are confirmed by Anthropic's own post. The company says the US government applied export controls to Claude Fable 5 and Claude Mythos 5 on June 12. Because Anthropic could not reliably verify nationality in real time, it suspended access for all users.

Anthropic says the controls were lifted on June 30. Fable 5 is returning globally to Claude Platform, Claude.ai, Claude Code and Claude Cowork. For Pro, Max, Team and select Enterprise plans, it is included for up to 50% of weekly usage limits through July 7, then moves to usage credits. Anthropic says access through AWS, Google Cloud and Microsoft Foundry will be restored as quickly as possible.

Mythos 5 access has also been restored for selected US organizations after government approval on June 26. The broader Glasswing partner rollout remains gradual.

A political scare became a product constraint for security work

The incident began with an Amazon researcher report that, according to Anthropic, found a way to bypass Fable 5 safeguards and get the model to identify software vulnerabilities. In one case, the model allegedly produced exploit demonstration code. Anthropic says the same vulnerabilities or the same demonstration could also be produced by less capable models including Opus 4.8, GPT-5.5 and Kimi K2.7.

The important part is that this did not end as a blog argument. It became a stricter classifier. Anthropic says the classifier blocks the reported technique in over 99% of cases, but also admits a higher rate of false positives for routine coding and debugging. Some requests will fall back to Opus 4.8.

For security teams, that is the practical cost. A model can become safer against misuse and worse for defensive work, because the queries often look similar.

The new classifier fixes the incident, not the rulebook

Zvi's core criticism is that US policy is still ad hoc. That point lands. If one report plus political escalation can shut a frontier model off for everyone, labs will optimize not only safety systems but also their relationship with government.

Anthropic wants a shared industry framework for assessing jailbreaks. That is the right direction. The weak point is precision. A stricter classifier can make a good slide for an official, but production teams care how many legitimate requests it drops and how many users move to weaker, cheaper or less governed models.

The refusal to usefulness ratio is now the metric to watch

The next signals are concrete: how quickly Anthropic reduces false positives, when cloud channels return, and whether the industry gets a jailbreak reporting standard that does not depend on a call to Washington.

Fable 5 is back, but the incident will be paid for in rejected debugging requests. If this becomes the default pattern, safety policy will shape developer workflow as directly as model quality.

Lilith's verdict

Fable 5 came back like a traveler released after an unnecessary airport search: the suitcase is taped shut, everyone claps and only at the gate do we learn how much useful gear security threw away.