Anthropic says its most powerful AI cyber model is too dangerous to release publicly — so ...

Anthropic launched Project Glasswing, a $100 million AI cybersecurity initiative using its unreleased Claude Mythos Preview ...