"In SWE-chat, vibe-coded commits introduced Semgrep security findings at approximately 3.8 times the human-only rate and 3.2 times the collaborative rate per thousand added committed lines."
The October 2026 study compares newly introduced static-analysis security findings across coding modes, measuring 0.47 findings per thousand added lines for vibe coding, 0.12 for human-only coding and 0.15 for collaborative coding. These Semgrep findings document an observed difference in security-relevant code patterns across the sampled workflows, rather than confirmed successful exploits.
View Source