Anthropic's Claude Mythos AI: Public Release, Restricted Access

Anthropic has released a new version of its advanced AI model, Fable 5, to the public. This model is part of its Mythos class, previously restricted due to cybersecurity concerns. Here's the thing: Fable 5 is designed for tasks like writing software code, answering complex research questions, and analyzing images. What's interesting is that Anthropic is restricting its use in sensitive areas. For example, most cybersecurity or biology and chemistry queries to Fable 5 will be routed to a less capable model, Opus 4.8. This also applies to attempts to extract the technology for training competing AI models in authoritarian countries. The company has hired outside experts and run a bug bounty program to test these restrictions. They say no one has found a way to completely unlock the model. Meanwhile, an unrestricted version, Claude Mythos 5, is available to companies and organizations already using this model family, including those in its Project Glasswing program. The bottom line: This release highlights the ongoing challenge of balancing advanced AI capabilities with security and responsible deployment.