OpenAI reports that its researchers [identified 1,724 vulnerabilities](https://winaero.com/948-linux-kernel-vulnerabilities-with-working-exploits-were-discovered-with-ai/) in the Linux kernel. They developed working exploits for 948, which could let local users gain root access. The team also found 291 vulnerabilities that could expose information.

Researchers tested promising findings in isolated virtual machines to filter out false alarms. They then refined working exploits, created potential fixes, and had people review each exploit and patch.

_These days, everything is vulnerable. Regardless of being created by humans or AI._