OpenAI's AI model found 948 Linux flaws and 1,724 vulnerabilities
OpenAI's unreleased AI model just found 948 security flaws hiding in the Linux operating system, all before hackers could get to them.
The project scanned a massive 2.8 million lines of code, pinpointing thousands of risky spots and confirming 1,724 real vulnerabilities, including ones that could let attackers gain extra access or leak sensitive information.
Human reviewers validated OpenAI findings
What's cool is the AI didn't just spot these issues: human security researchers reviewed the results before reporting vulnerabilities and proposing fixes.
This approach turned out to be way more budget-friendly than relying on humans alone (think up to 55 times cheaper).
Through its "Patch the Planet" program, OpenAI worked with Linux maintainers and partners under its Patch the Planet initiative to report and address additional flaws, showing how powerful AI can be for keeping our technology safer and more affordable.