BREAKING: Fable and Mythos have been taken down for security concerns.
TL;DR
The shutdown was immediate and sweeping: Anthropic says it received the government directive at 5:21 p.m. Eastern and had to disable Fable 5 and Mythos 5 for all customers because it could not yet verify citizenship status.
Foreign nationals are the specific target: The order blocks access by any foreign national, whether inside or outside the US, and even applies to foreign national Anthropic employees, which Theo calls an unbelievable precedent.
Anthropic argues the jailbreak was narrow and not unique: According to the statement, the government showed a method that could get Fable to identify minor software vulnerabilities, but Anthropic says other public models, including OpenAI's GPT-5.5, can do the same without a bypass.
Fable was already heavily locked down: Theo notes Anthropic had strong cyber safeguards, thousands of hours of red-teaming with the US government and UK AISI, and a new 30-day prompt retention policy that already cost them customers.
Theo thinks the policy standard is dangerous: He highlights Anthropic's warning that recalling a commercial model over a non-universal jailbreak would effectively halt new frontier model deployment if applied consistently across the industry.
This instantly makes Anthropic's policy writing look prescient: Theo points to Dario Amodei's recent essay on AI's strategic consequences and says the company just got the "fast reaction" from government it had been warning about.
The Breakdown
Anthropic abruptly pulled Fable 5 and Mythos 5 after a Friday evening US export control order, and Theo is stunned that a narrow jailbreak report was enough to block access for all foreign nationals, including Anthropic employees. His big point is that if this standard holds, frontier model rollouts across the industry could grind to a halt.
Was This Useful?
Share
Keep Reading
Make Alcreon Yours
Tune your feedFive quick questions, and the feed ranks what matters to you first.Or just get notified
The weekly Echo. Signal worth keeping in your inbox.
Every new piece, announced on X.
Read Next
See all
Playbook
Cheap Models, Hard Tasks
Most agent workflows route every step to the frontier model by default. The bill scales with how chatty the agent gets, even when most steps don't need that brain.

Playbook
Tasteful Skills
“Tasteful Skills” argues that the best agent skills are not documentation or best-practice lists.

Playbook
The Art of Tasteful Prompting
Learn how tasteful prompting helps you move beyond generic AI output by shaping context, style, and judgment from the start.