OpenAI’s autonomous cybersecurity system identified 948 distinct privilege-escalation vulnerabilities in the Linux kernel and developed 1,169 working exploits. Its agents analyzed roughly 2.8 million lines of code, about 7% of the project, and dynamically reproduced 1,724 candidate vulnerabilities. The October run cost $1.21 million in API credits, or about $1,279 per vulnerability, compared with 166 unique flaws found in July by an earlier model version. The system also drafted candidate patches, but human researchers and software maintainers remain responsible for validating and implementing them. OpenAI says automating exploitation could help defenders patch the kernel faster, while also making outdated operating systems easier to compromise.
Source: Read the original article
