← Back to feed

The US Department of Commerce has lifted export controls on both Claude Fable 5 and Mythos 5. Anthropic confirmed late Tuesday that it received formal notice, with commercial access restoring Wednesday July 1.

The controls were imposed June 12, when Anthropic abruptly disabled both models after receiving a directive from Commerce. The shutdown lasted 19 days and cut off more than 200 institutions, including universities, government agencies, and enterprise customers, that had been relying on the two highest-capability models in Anthropic’s lineup.

What Changed

The full restoration covers both models simultaneously. Fable 5, the publicly available frontier model posting 95.0% on SWE-bench Verified, returns to commercial availability for all customers. Mythos 5, the higher-clearance variant at 95.5% SWE-bench Verified, resumes access for the 100+ partner organisations that had been cleared in the limited interim unlock.

The partial unlock of Mythos in the weeks following the ban, which covered critical infrastructure defenders and a subset of US companies, did not include Fable 5. That asymmetry ends today.

The Saga in Brief

The ban originated from a Commerce Department directive that applied export controls to both models simultaneously. Anthropic did not contest the controls publicly at the time, though multiple reports indicated the company pushed back through government channels on the scope and duration.

Speculation around a resolution ran for two weeks. Markets were pricing a 67% probability of a return by July 1 in the days before the announcement. The June 30 confirmation hit that deadline precisely.

The G7-connected policy track that emerged mid-saga, which proposed a trusted-partner access framework for future frontier model gating, remains in discussion but was not a precondition for this restoration.

The Models Returning

Claude Fable 5

• Intelligence Index: 64.9 (Artificial Analysis, ranked #1)
• SWE-bench Verified: 95.0%
• SWE-bench Pro: 80.3%
• Terminal-Bench 2.1: 84.3%
• Agent Arena: #1, 12.94% composite

Claude Mythos 5

• SWE-bench Verified: 95.5%
• Terminal-Bench 2.1: 88.0%
• AISI cyber range: cleared all 32 steps

Mythos 5 cleared AISI’s 32-step corporate cyberattack range in full, which was the original trigger for the export control action: a capability assessment that determined the model posed potential dual-use risk in offensive cybersecurity contexts.

What Comes Next

Anthropic has not published new terms for Fable 5 access following the restoration. The existing enterprise zero-data-retention agreements that Fable 5 voids for safety surveillance purposes remain in effect.

The Commerce Department has not published a formal explanation for the lifting. The White House AI legislative framework, released separately in June, proposed a customer-by-customer approval mechanism for future frontier model launches of this capability tier. Whether that framework was satisfied during the negotiations or remains pending as a parallel track is unclear.

For the 200+ institutions that lost access on June 12, the restoration closes a three-week gap in their most capable AI deployments. The costs of that gap, across research pipelines, enterprise deployments, and security programs, have not been quantified publicly.