Daily Briefing: Anthropic Ships Its Most Capable Model Twice

Anthropic simultaneously released Claude Fable 5 to the public and Claude Mythos 5 to US government cyberdefense partners through Project Glasswing. Both are built on the same architecture, but Fable 5 includes automatic safety classifiers that silently route dangerous queries to the less capable Opus 4.8 model. The dual-release establishes a new template for frontier AI distribution: guardrailed access for commercial users, unrestricted access for vetted security partners. Controversy has already emerged around Anthropic's 'hidden interventions' that quietly degrade model performance on certain AI development tasks without notifying users, creating a tension between the company's safety commitments and its transparency brand.

STORIES COVERED

Anthropic ...