Suspected AI-generated code gets GitHub PRs rejected with shallow feedback, study finds

soumitrashukla9 · x · 2026-09-17

A job-market paper by Berkeley PhD student Sahiba Chopra built algorithms to detect potential AI use in GitHub PRs and found moderators are more likely to reject PRs and give generic, shallow feedback when they suspect code is AI-generated — revealing a hidden bias against AI-assisted contributors in open source.

Original post →

More from Research

Research channel →