OpenAI Reports 948 Linux Kernel Flaws Found by AI

AI · First reported 5 Nov, 15:15 · Updated 18:55

Image: Wikimedia Commons · NASA Johnson Space Center · Public domain

OpenAI says its cybersecurity system identified 948 distinct Linux kernel privilege-escalation vulnerabilities and produced working exploits. The experiment used about $1.21 million in API credits; OpenAI says it also generated candidate fixes, which still require human review and implementation.

OpenAI says its agents found 948 distinct Linux kernel privilege-escalation vulnerabilities and produced 1,169 working exploits.
The October run scanned about 2.8 million lines of kernel code and cost $1,213,430 in API credits, according to OpenAI.
OpenAI says the system drafted candidate fixes, but human reviewers and software maintainers remain responsible for validating and implementing them.

Summary written by Lux AI from 2 articles

Key facts

Distinct vulnerabilities: 948 Linux kernel privilege-escalation flaws, according to OpenAI.
Working exploits: 1,169 reported.
Code scanned: About 2.8 million lines, or roughly 7% of the project.
Reported API-credit cost: $1,213,430 for the October run.