FULL STORY
AI-Generated Code Security Pass Rate Improves for First Time
Veracode's GenAI code security report showed AI-written code passing security checks at just 56%; its fall benchmark update marked the first significant improvement, rising to 62%.
2026-10-11 ~ 2026-10-30 · 2 episodes · 6 posts
Episode 1 · AI Now Writes Half of Code but Passes Security Checks Only 56% of the Time (2026-10-11, 2 posts)
Veracode's 2026 GenAI Code Security Report finds that AI now writes about half of committed code in teams that have adopted it, yet testing across 100+ models shows an average security pass rate of just 56%, meaning roughly 44% of AI-generated tasks contain known vulnerabilities.
- Veracode: AI now writes half of all code, but security pass rate stuck at 56% — WeldPond · 2026-10-11
- AI writes half of committed code but only 56% of models pass security tests — Thionne_WTZ · 2026-10-13
Episode 2 · AI Code Security Pass Rate Rises for First Time: Veracode Benchmark (2026-10-30, 4 posts)
Veracode's GenAI code security benchmark showed its first significant improvement, with the average security pass rate rising from 56% to 62% and a new record of 73% shared by Claude Opus 5.5 and another model.
- AI Code Security Finally Improves: Veracode Benchmark Pass Rate Jumps From 56% to 62% — WeldPond · 2026-10-30
- LLMs finally write more secure code: Veracode benchmark average jumps 56% to 62% — WeldPond · 2026-10-30
- Veracode benchmark improves for first time: LLM code security pass rate jumps 56% to 62% — WeldPond · 2026-10-30
- Veracode benchmark: LLM code security pass rate jumps from 56% to 62%, first-ever gain — WeldPond · 2026-10-30