Study from 1Password's Off-By-1-Labs finds AI produces correct patches 26%
Technology
Turns out, AI still has a long way to go when it comes to fixing software bugs.
A fresh study from 1Password's Off-By-1-Labs found that AI tools like Claude and Codex only got security patches right 26% of the time across six recent vulnerabilities, including CVE-2026-31431.
Even worse, over half the patches either didn't work or actually made things messier, which really shows why human experts are still crucial for cybersecurity.
Researchers launch flawed tool on GitHub
To tackle this, the researchers have launched a new tool called FLAWED on GitHub.
It's designed to help researchers and companies figure out where AI can safely handle patching, and when you definitely need a human in the loop.
The goal? Keep improving how AI helps keep our tech safe.