Google Cloud launches AI Threat Defense for automated vulnerability patching

google cloud unveiled ai threat defense, a security platform integrating technologies from deepmind, cloud security firm wiz, and mandiant to automate the detection and remediation of software flaws. the system employs multiple ai models to scan cloud infrastructure and code, utilizing cheap models for continuous checks and frontier models for critical systems. deepmind's codemender agent enters the development environment to rewrite vulnerable code into memory-safe languages and generates automated tests to verify patches before deployment. this launch responds to the evolving threat landscape where ai models like anthropic's claude mythos can discover vulnerabilities and generate exploits within minutes, rendering traditional manual patching and 90-day grace periods obsolete. partners include accenture, deloitte, and pwc, though the reliability of autonomous code generation in production environments remains a key concern.