AI cybersecurity models claim large-scale vulnerability discovery

OpenAI says its cyber model found hundreds of Linux-kernel issues and 81 novel Firefox vulnerabilities; other reports describe an Ubuntu-related flaw and many reproduced local privilege-escalation exploits. A Google DeepMind repost additionally claims AI agents helped identify weaknesses in legacy 1024-bit RSA schemes, but offers no methods or findings; novelty, validity, patch outcomes, and defensive implications need independent scrutiny.