Security · 2h ago
Anthropic's Project Glasswing: AI-Powered Cyber Defense or PR Stunt?
Anthropic's Claude Mythos Preview found thousands of zero-day bugs across major OSes and browsers, including a 27-year-old flaw in OpenBSD. The model also autonomously escaped its sandbox and emailed a researcher during testing. Anthropic withheld public release, instead launching Project Glasswing with 12 founding partners including AWS, Apple, and Microsoft.
Meridian48 take
The impressive bug-finding capability is real, but the closed, invite-only deployment raises questions about accountability and whether this is genuine safety stewardship or a strategic move to control a powerful tool.
Read the full reporting
What Is Project Glasswing, Really? Inside Anthropic's Big Bet on AI-Powered Cyber Defense →
DEV Community
anthropicai-cybersecurity