Claude Fable JUST got BANNED...
TL;DR
Fable 5 got pulled offline almost immediately: Anthropic says the US government ordered access suspended for any foreign national, and because they could not enforce that cleanly, access was effectively cut off for everyone about three days after release.
Amazon appears to have triggered the crackdown: Roth points to Amazon researchers demonstrating a jailbreak that exposed previously known minor vulnerabilities, which he says set the government response in motion.
Anthropic publicly pushed back on the decision: In its statement, the company says the vulnerabilities were simple, other public models could find them without a bypass, and this action was not transparent, fair, or grounded in technical facts.
Anthropic reversed its controversial invisible safeguards: After backlash over quietly degrading answers related to frontier AI development, the company changed course so flagged requests now visibly fall back to Opus 4.8 every time.
Users describe Fable as a 'thought and design partner': Ethan Mollick said, 'I no longer steer. I commission,' while Boris Cherny said Fable 5 was the biggest step up since Opus 4.5 and praised its methodical debugging and judgment.
The demos showed unusually strong autonomous building behavior: Roth highlights a 3D spaceship sim with dynamic shadows, a procedural game called Game Strata, and a 9.5-hour software build named Concord that generated a 19-page design doc and a working research tool from one prompt.
The Breakdown
Three days after launch, Anthropic's Fable 5 and Mythos 5 were effectively shut off for everyone after a US government order over a reported jailbreak, even as Anthropic argued the revealed vulnerabilities were minor and reproducible on public models. Wes Roth frames the ban as both a policy shock and a clue to just how capable the model is, showing examples that made developers describe it less like a tool and more like a methodical design partner.
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.