South Korean Financial Institutions Hit by Suspected AI Agent Breaches; Anthropic's Mythos AI Discovers Critical Zero-Day

South Korea's financial regulator is investigating coordinated intrusions at seven financial firms, believed to be automated by AI agents. Separately, a critical authentication bypass vulnerability (CVE-2026-61500) in Rejetto HTTP File Server, discovered by Anthropic's access-restricted Mythos AI model, is now being exploited in the wild, granting full admin access and remote code execution.