Matthew BermanJune 13, 202615m

WTF is going on?!

TL;DR

Anthropic pulled the plug in under 3 hours: After receiving a government directive at 5:21 p.m. Eastern, Anthropic disabled Fable 5 and Mythos 5 for all non-US citizens, and Berman says his own 10 active agents stopped mid-run almost instantly.
The definition of 'foreign national' makes this wildly broad: Berman notes it appears to cover basically any non-US citizen, including green card holders and even Anthropic employees, which raises obvious enforcement questions for an API business unless it starts doing bank-style KYC.
Berman says this was a self-inflicted wound: He traces the ban back to Anthropic's Project Glasswing and Mythos rollout, where the company repeatedly framed its own model as uniquely dangerous in cyber, bio, and nuclear domains, then released Fable 5 as a filtered version of that same model.
Amazon helped push the security alarm: Citing The Wall Street Journal and The Information, Berman says Amazon researchers demonstrated a jailbreak and Amazon CEO Andy Jassy raised concerns to Trump administration officials, despite Amazon being one of Anthropic's biggest investors and vendors.
The core technical trigger looks ordinary, not exceptional: Anthropic's own response says the jailbreak exposed a small number of previously known minor vulnerabilities, and Berman's point is blunt: every frontier model can be jailbroken, so this does not make Fable or Mythos uniquely dangerous.
This could delay Anthropic's IPO and shift how AI is treated: With Anthropic having confidentially filed its S-1, Berman thinks the bigger consequence is not buzz but being recast from a product company into a national security risk, which could harden regulation across the whole AI industry.

The Breakdown

Anthropic shut off Fable 5 and Mythos 5 for all foreign nationals within hours of a US export control directive, after years of telling everyone these models were too dangerous to release freely. Matthew Berman argues the ban was triggered by jailbreak concerns every major model shares, and that Anthropic's own fear-heavy messaging helped turn its flagship product into a national security problem.