Anthropic's cybersecurity system flags over 10,000 vulnerabilities in weeks, highlighting a growing gap between rapid AI-driven discovery and slower patching and disclosure processes