Short answer: The United States applied export controls to Anthropic’s Claude Fable 5 and Claude Mythos 5 on 12 June 2026, and lifted them on 30 June. Anthropic suspended both models globally because it could not verify users’ nationality in real time, then restored Fable 5 worldwide on 1 July after updating its safeguards.
Last reviewed: 28 July 2026. This explainer distinguishes Anthropic’s confirmed account from details reported by news organisations.
AI export-control timeline
| Date | What happened | Confirmed by |
|---|---|---|
| 9 June 2026 | Anthropic released Claude Fable 5 and made Mythos 5 available to approved Project Glasswing partners. | Anthropic |
| 12 June | The US government applied export controls requiring access restrictions for foreign nationals. Anthropic suspended both models for all users because it could not verify nationality instantly. | Anthropic |
| 26 June | The government approved restored Mythos 5 access for a limited set of US organisations. | Anthropic |
| 30 June | The controls on Fable 5 and Mythos 5 were lifted. | Anthropic |
| 1 July | Fable 5 returned globally. Mythos 5 remained limited to approved organisations while Anthropic worked to broaden partner access. | Anthropic |
Why did the US restrict the Claude models?
According to Anthropic’s incident account, the directive followed an Amazon research report describing a way to bypass Fable 5’s safeguards. The technique prompted the model to identify software vulnerabilities and, in one case, produce code demonstrating how a vulnerability could be exploited.
Anthropic said its own testing found that less capable models could identify the same vulnerabilities and that the reported behaviour did not reveal unique Mythos-level capabilities. That is Anthropic’s assessment, not an independently published government finding. The underlying directive was not linked in the public material reviewed for this update.
Why did the restriction cause a global shutdown?
The order required Anthropic to prevent foreign nationals from accessing the affected models, including people located inside the United States. Anthropic said it had no reliable real-time system for verifying every user’s nationality, so it suspended access for everyone to comply immediately.
This distinction matters: the government did not announce that every domestic user was prohibited from using the models. The wider outage was Anthropic’s operational response to a restriction it could not enforce selectively at short notice.
What changed before access returned?
Anthropic trained an improved safety classifier aimed at the technique described in the Amazon report. The company says the classifier blocks that specific bypass in more than 99% of cases, although it can also flag some benign coding and debugging requests. Researchers at the US Department of Commerce’s Center for AI Standards and Innovation tested the old and new safeguards, according to Anthropic.
After the controls were lifted on 30 June, Fable 5 returned globally the next day. The sequence suggests that the government and Anthropic treated the issue as a specific safeguard and access-control problem rather than a permanent finding that the public model was too dangerous to export.
What the reversal means for AI governance
- Model access can change faster than account systems. A nationality-based rule is difficult to implement when consumer AI services normally verify location or payment details rather than citizenship.
- Safeguards are now part of market access. Classifier performance and jailbreak remediation can affect whether a frontier model remains available.
- Public evidence remains incomplete. Anthropic published a detailed account, but the full government directive and independent test results were not included in the public sources reviewed here.
- One incident is not a general AI-export framework. The episode shows how authorities handled these two models; it does not by itself establish a durable rule for every future model.
How this connects to the July 2026 model releases
The restriction occurred during a fast sequence of model and pricing changes. Anthropic also made Claude Sonnet 5 the default model for Free and Pro users, while Google expanded its lower-cost Gemini image-generation model family. Readers comparing these releases should check current availability, safeguards and per-output pricing—not just headline model names.
For a broader framework, use our 12-point AI tool evaluation checklist before moving a team or workflow to a newly released model.
Why was the AI model banned?
Why was the ban reversed?
How long did the ban last?
Does this mean AI regulation is working?
Primary and supporting sources
- Anthropic: Redeploying Fable 5 — timeline, access impact, safeguard changes and restoration.
- Anthropic: Claude Fable 5 and Claude Mythos 5 — model release and access design.
- NPR: reporting on the restrictions.
- Nextgov/FCW: reporting on the controls being lifted.
Update log: 28 July 2026 — added the operational distinction between the government restriction and Anthropic’s global shutdown, replaced unsupported inference with attributed claims, added a dated timeline, primary sources and links to the related Claude and Gemini explainers.