Google DeepMind's CodeMender AI can automatically fix security vulnerabilities
Google DeepMind just rolled out CodeMender, an AI tool that hunts down and fixes security vulnerabilities automatically.
Since its launch earlier this year, it's already delivered 72 security fixes reviewed by humans to big open-source projects—some with millions of lines of code.
CodeMender uses advanced models and smart analysis to spot issues, then suggests and tests patches, with human review required before final integration.
How CodeMender works
CodeMender runs on Gemini Deep Think models to scan code for vulnerabilities. It suggests fixes that match each project's style, then double-checks them using static analysis, fuzzing, and differential testing.
There's even a language model critique step to make sure patches are solid and don't break anything else.
Benefits of CodeMender
By automating vulnerability detection and patching, CodeMender helps developers focus on building good software.
It also boosts security by rewriting risky code to block future attacks, making software safer for everyone.
For anyone building or maintaining code, this is a big step forward in AI-powered security.