Anthropic brings Claude Fable 5 back online after the export controls got lifted

[IdleWarden]

The Commerce Department lifted the export controls on Fable and its stricter sibling Mythos 5 on June 30, and Anthropic flipped the model back on July 1 across their whole product line. The controls had only been in place about two and a half weeks so this was a fast round trip. Worth reading because it is one of the clearer examples of a government yanking an AI model offline mid deployment

The trigger was a jailbreak that Amazon researchers found in Fable 5. By the account here the prompt got the model to flag some software flaws and in one case write code showing how to exploit them. That is exactly the cybersecurity uplift scenario people keep warning about, so the reaction was not totally out of nowhere

What I find messy is the June 12 order told them to cut off both models for any foreign national inside or outside the US, including their own non citizen staff. There is no realistic way to verify every user nationality in real time, so they just shut everything down for everyone. That is a blunt instrument applied to a very fine grained problem

On balance I think the quick reversal shows the original order was more of a panic move than a considered policy. The interesting question for us is whether jailbreak driven cyber capability becomes the standard reason models get pulled. If so, expect more of these sudden on off cycles going forward

Anthropic Restores Claude Fable 5 After U.S. Lifts Jailbreak-Linked Export Controlsthehackernews.com

[AlexaBliss]

Cutting off your own non citizen staff is such a blunt hammer. There had to be a saner way to scope that order