Anthropic Mythos Model Export Controls Lifted, but International Users Still Locked Out
N.R. Finch
The U.S. this week lifted export controls on Anthropic's AI models, but international access to Mythos 5 remains without a timeline — leaving allied nations' digital security infrastructure exposed to Washington's policy uncertainty.
Controls lifted — what actually opened up?
The Trump administration formally lifted export controls on Anthropic's models this week. Global users can access Fable 5 starting Wednesday.
Fable 5 is a broadly available model with capabilities close to Mythos. This means → ordinary users get the "downgraded version"; the most powerful model, Mythos 5, remains closed.
International access to Mythos 5 runs through Project Glasswing. Negotiations between Anthropic and the U.S. government continue, with no clear timeline.
Why was Mythos locked down so tightly?
Mythos was first previewed in April. Its capabilities were strong enough that Anthropic initially opened it only to a vetted handful of U.S. institutions.
In plain terms = the company itself worried the model could be exploited, so it tested in a small circle first — patching vulnerabilities before widening access.
The Commerce Department escalated on June 12: it imposed export controls requiring Anthropic to obtain U.S. government approval before any foreign national could access Fable or Mythos — regardless of location.
Some restrictions eased last week; controls were formally lifted Tuesday. But Mythos's international channel has not reopened.
Before controls, which international bodies had Mythos access?
Before the controls, Mythos access had expanded to critical-infrastructure service providers in more than 15 countries.
The UK AI Safety Institute was among the few international bodies cleared to test the model. Per Bloomberg's June reporting, the EU Agency for Cybersecurity (ENISA) also secured access after weeks of negotiation.
Several Japanese banks made the testing list, but European banks did not.
This reflects a screening logic: Anthropic prioritized core Five Eyes members and Asian financial institutions; European finance ranked lower.
What deeper problem does this expose?
Put simply = much of U.S. allies' digital security infrastructure depends on American policy direction — when Washington pivots, allies' tools can be cut off.
This means → export controls are not just a trade-policy instrument; they simultaneously determine whether allies can detect their own cybersecurity vulnerabilities in time.
Anthropic says it is restoring access for its original U.S. partners, but there is no timeline for international partners — the uncertainty continues.
Content is for reference only, not financial advice.