On 30 June the export controls imposed around 12 June on Claude Fable 5 and Mythos 5 fell away. Access is coming back - in exchange for a deal that Anthropic will monitor and report the risks.
- On 30 June the US lifted the export control on Claude Fable 5 and Mythos 5, imposed around 12 June.
- The deal: Anthropic proactively monitors and reports security risks and works with the government on standards.
- Access is restored through Claude, claude.ai and Claude Code, then also on AWS, Google Cloud and Azure Foundry.
A resolution from June. On 30 June the US Department of Commerce lifted the export control on Claude Fable 5 and Mythos 5 (code names for new Claude models). The freeze started around 12 June - over national security and a reported jailbreak-type vulnerability - and cut off access for foreign nationals, while the models were switched off for everyone. Now the control falls away, with no export license needed.
This makes me ask about the precedent, not the resolution. An entire government cut off access to a powerful AI model for days over one report, then restored it after negotiations. The speed of the reversal says more about political pressure than about a settled technical dispute.
For those of us building on these models, the lesson isn't panic. The lesson is dependency. If the tool can be stopped with one order, resilience is called a backup provider. It came back - good. But remember how it disappeared, if you're building on it.