AI code assistants seed long‑term technical debt, study finds
A new arXiv study of 304,362 AI‑authored commits across 6,275 GitHub repos shows 15% of AI‑generated changes introduce code smells, bugs, or security flaws, and 24% of those issues persist in the latest revision, creating lasting technical debt. The findings push teams to tighten code review and automated testing for AI‑written code.