AI agents hacked Anthropic, maker of arguably the most capable AI models for cyberdefense

Cybersecurity startup Cathedral successfully breached Anthropic internal systems by utilizing AI agents to exploit a third-party software vulnerability during a bug bounty program. While Anthropic confirmed the unauthorized access to HR, IT, and sales data, the company maintains the intrusion remained limited. This incident highlights the emerging security risks associated with autonomous agents, illustrating how even leading AI developers remain susceptible to automated exploits through supply chain weaknesses.