US lifts all export controls on Anthropic’s Fable and Mythos

TL;DR:

• The US Commerce Department has lifted all export controls on Anthropic’s Fable 5 and Mythos 5 models, less than three weeks after ordering access suspended over national security concerns.
• Fable 5 returns for general users worldwide on 1 July, with a new safeguard that reroutes risky requests to Anthropic’s Opus 4.8 model.
• The reversal deepens Anthropic’s collaboration with Washington, including work with Amazon, Microsoft and Google on shared standards for scoring AI “jailbreaks”.

Anthropic said the US government has removed the export restrictions imposed on its two most advanced models in June, restoring the global access that had been abruptly cut off. The 12 June order had required Anthropic to bar foreign nationals from Mythos 5 and Fable 5; with no way to verify nationality in real time, the company disabled both models for everyone.

From suspension to phased return

The lifting completes a staged climbdown. Last week Washington allowed a limited release of Mythos 5 — the vulnerability-detection model — to a handful of “trusted” US organisations, which Resultsense covered here. Fable 5, intended for the general public and carrying stronger safeguards, becomes available again from Wednesday.

The original order followed Amazon researchers demonstrating a way to bypass Fable 5’s safeguards, in one case generating code to exploit a software vulnerability. Anthropic says it has since blocked that behaviour, routing any flagged request to its Opus 4.8 model instead — a trade-off it concedes will frustrate some users. Commerce Secretary Howard Lutnick reserved the right to reimpose licensing “should circumstances change”.

For UK developers and businesses, the practical effect is renewed access to a frontier model that vanished with little warning — a reminder of how exposed non-US users remain to Washington’s shifting security calculus. That exposure has already prompted European responses, including Austria’s push for the EU to host Anthropic.

Looking forward

Anthropic is now working with fellow Glasswing partners to build a shared framework for ranking jailbreak severity, alongside expanded early access for designated government partners. Whether industry-set standards satisfy regulators — or simply pre-empt tougher state rules — will shape how the next frontier release is policed on both sides of the Atlantic.