AI

OpenAI's AI Found 948 Ways to Break Into Linux — and Wrote the Fixes

📅 Nov 5, 2026

✍️ Aamir Raza

OpenAI's AI Found 948 Ways Into Linux

One automated run turned up nearly a thousand paths to root — and wrote a patch for every one.

LexxWire

From 2,308 Leads to 948 Root Exploits

The pipeline checked 2,308 candidate flaws, confirmed 1,724 as real, and built 948 working root exploits with patches.

LexxWire

The Whole Run Cost $1.2 Million

About $1,279 per exploit — OpenAI says the same haul would have cost $19M to $66M in bug bounties.

LexxWire

It Found 10 Tricks No Expert Had Seen

The agents used at least 10 new exploitation techniques, including new ways to spray kernel memory and reach root through data alone.

LexxWire

Fixes Are Already Landing

61 flaws were patched upstream within days, 43 more had patches in staging, and hundreds of reports went to kernel maintainers.

LexxWire

More Stories

Apple's New CEO Wants Siri to Become a Real AI Agent

Nvidia Is Investigating Its Own CEO, Jensen Huang

Read Full Story →

Thanks for reading!

Next: Apple's New CEO Wants Siri to Become a Real AI Agent