Anthropic AI model helps find 3.5 times more security bugs

A new report from cybersecurity firm Epoch AI shows that in June 2026, 21 organizations reported about 1,500 high-severity and critical security bugs. That's more than 3.5 times the previous monthly record. The trend started in April, linked to Anthropic's Claude Mythos Preview model.

This model is Anthropic's strongest bug-finding AI. Before its official release, trusted partners were already using it to find and fix vulnerabilities. Anthropic's 'Glasswing' project has found over 10,000 high-severity or critical bugs so far, some still undisclosed.