The US Government Forced Anthropic to Pull Fable 5 and Mythos 5

TL;DR

• 1

  The US government forced Anthropic to shut off Fable 5 and Mythos 5: Anthropic says an export control directive barred access for foreign nationals, but because it could not reliably filter by citizenship, it disabled the models for everyone.

• 2

  A jailbreak report from Amazon appears to have triggered the crackdown: Reporting cited by the hosts says Amazon researchers found a way to surface more dangerous cyber-related behavior, and Andy Jassy may have escalated it to administration officials despite Amazon having invested up to $33 billion total in Anthropic.

• 3

  Dario Amodei had just argued for exactly this kind of government power: In his June 10 essay, he said frontier models should face testing, auditing, and even blocked or reversed releases if they pose public safety risks, while also warning against arbitrary or politically motivated decisions.

• 4

  The real shock is that public models are not the strongest models in the lab: The hosts frame Fable 5 as a guarded public version of the more powerful Mythos system, a reminder that providers may already be withholding capabilities far beyond what most users see.

• 5

  Businesses now have to treat model access as a strategic dependency risk: If Anthropic, OpenAI, or Google models can be restricted by pricing, compute limits, or government action, companies may need to control more of their own stack, possibly pushing interest back toward open source.

• 6

  This may be the start of government claiming first rights to frontier AI compute and capability: The hosts speculate that agencies could demand review access, reserve chunks of compute for national security use, and shape what gets commercialized, creating uncertainty for product roadmaps and future IPOs.