GoKawiilCalvin Wankhede / Android Authority
TL;DR Anthropic introduced its Fable 5 and Mythos 5 models last month, with advanced security research tools.
Even though Anthropic took major efforts to limit abuse potential, it still pulled the models following a government directive.
Fable 5 and Mythos 5 access has now been restored, and Anthropic is explaining exactly what happened.
It has been a whirlwind of a month for Anthropic and a couple of its latest model launches: Claude Fable 5 and Mythos 5. First introduced three weeks ago, the models promised such powerful cybersecurity capabilities that Fable had to be constructed as an essentially locked-down version of the access-restricted Mythos, designed to block attempts to use it for malicious purposes. But even with those precautions, a government export control directive forced Anthropic to pull the plug on both. We heard they’d likely be back soon, and today, Anthropic formally shares word of the return of Fable 5.
As of today, July 1, Anthropic announces that Fable 5 is once again available for use — and that’s globally, as well, with no US-only restriction.
The company shines a little light on the abrupt rug-pulling of these models, and to hear its explanation, the government’s concern extends from a big misunderstanding. Amazon researchers reportedly found a way to side-step some of Fable 5’s safeguards, and this apparently sparked the export-control action. But upon closer inspection, the results those researchers got from Fable 5 were not directly tied to its advanced hacking capabilities, and could have also been produced by other generally available models, like OpenAI’s GPT-5.5.
While the government’s order at the time only technically applied to foreign access to the models, without a robust system in place to authenticate nationality, Anthropic felt it had no choice but to simply flip the switch off for everyone.
Even taking all that into account, there’s still a lot of damage control going on with this statement, and Anthropic spends a fair amount of time outlining how committed it is to not just hardening Fable 5 against jailbreaking efforts to override its security model, but also how it plans to characterize and respond to such efforts going forward — something it says it’s working on with partners like Amazon, Microsoft, and Google.
Anthropic further emphasizes its willingness to share models with the government for pre-release preview, getting feedback on national security concerns.
... continue reading