Anthropic's Mythos 5 Clears US Export Controls for Authorized Cybersecurity Teams
After two weeks of negotiations with the Trump administration, Anthropic secured limited access to its flagship security model for approved domestic organizations.
Last verified:
Mythos 5 Returns Under Conditional Licensing
Anthropic secured selective access to Mythos 5, its strongest cybersecurity model, for approved domestic organizations following protracted discussions with the Trump administration. According to The Verge, Commerce Secretary Howard Lutnick issued a June 26 letter to Anthropic co-founder Tom Brown authorizing deployment to a restricted set of cyber defenders and infrastructure providers—a structure mirroring the concurrent approval granted to OpenAI for GPT-5.6.
The conditional clearance came after negotiations spanning two weeks, following an earlier administration order that blocked all foreign-national access to both Mythos 5 and its public-facing variant, Fable 5. Anthropic spokesperson Danielle Ghiglieri confirmed that the company had “received notice” permitting redeployment to the curated provider network and stated the organization is “working to provision the approved set of providers and restore their access.”
Fable 5 Remains Sidelined
Unlike Mythos 5, Fable 5—Anthropic’s intended general-use model—has no announced path to reauthorization. The Verge reports that no timeline exists for lifting restrictions on the public-facing release, leaving Anthropic’s consumer-facing offering in regulatory limbo indefinitely.
The blanket prohibition on foreign-national access remains in effect outside the approved cohort, meaning Anthropic employees who are not US citizens continue to be barred from either model unless they join an authorized organization.
Why This Matters
The carve-out establishes a precedent for conditional US AI access tiers: governments can now distinguish between closed and open deployments, permitting restricted domestic access while maintaining export controls on broader releases. This framework may shape how other frontier-model developers negotiate with regulators over cybersecurity-critical systems.
For Anthropic, the approval unblocks revenue from cybersecurity and infrastructure contracts but postpones consumer monetization of Fable 5 indefinitely. Competitors including OpenAI now operate under similar restrictions, meaning the market for unrestricted frontier models may remain segmented by nationality for the foreseeable future—a structural change that could reshape international AI expansion strategies.
Frequently Asked Questions
Did Anthropic win full export-control relief?
No. The Trump administration granted a carve-out for Mythos 5 access by approved organizations, similar to OpenAI's GPT-5.6 exception. Fable 5 remains under the blanket export restriction, and the foreign-national access ban remains in effect outside the approved list.
Which organizations can now access Mythos 5?
According to The Verge, the approved set includes cyber defenders and infrastructure providers, though the specific organization list has not been disclosed. Anthropic is provisioning access to this pre-approved cohort.
What was the original restriction?
Two weeks prior, the Trump administration barred all foreign nationals—including Anthropic employees—from accessing Mythos 5 and Fable 5, effectively suspending both models internationally.