Yes

>>A Look Inside Claude Mythos: Anthropic’s New Confidential Model and Its Approach to Transparency

A Look Inside Claude Mythos: Anthropic’s New Confidential Model and Its Approach to Transparency

crypthub

about 5 hours ago

Glasswing Initiative

Anthropic has launched the Glasswing project built around its new Claude Mythos Preview model. The effort is framed as a step toward deeper insight into advanced AI behavior. Rather than a commercial product, the model serves as a research platform. Anthropic stresses that interpretability is as important as raw performance.

Restricted Deployment

Mythos Preview will not be released to the public; access is limited to select tech giants such as AWS, Apple, Google, Microsoft, NVIDIA and others. The company says the restriction helps secure the world’s most critical software. Partners receive up to $100 million in usage credits for testing. Anthropic aims to balance innovation with safety by favoring analysis over rapid rollout.

Security Discoveries

Within Glasswing the model has already uncovered thousands of high‑severity, zero‑day vulnerabilities across major operating systems and browsers. Notably, it exposed a 27‑year‑old flaw in OpenBSD that could crash remote machines. These findings illustrate Anthropic’s shift from racing new capabilities to first understanding them. The project will publish lessons learned to inform broader industry practices.