Veracode Finds AI-Generated Code Security Has Barely Improved Since Last Year

Veracode's 2026 GenAI Code Security Report reveals that AI-generated code security has stalled at a 56% average pass rate across 100+ models — nearly unchanged from last year. Despite AI generating roughly half of all committed code, nearly 44% of AI-generated code fails security tests when no security-specific prompting is used. Coding-specialized models (51% pass rate) are no safer than general-purpose models (52%), and model size has no meaningful impact on security performance. Reasoning models show a slight edge at 56% vs. 51% for non-reasoning models. GPT-5.5 leads the leaderboard at 68%, while Java remains the riskiest language at only 30% pass rate. Veracode recommends integrating security scanning into AI-assisted development workflows and treating AI-generated code like any unreviewed code.

Nguồn: https://sdtimes.com/agentic-security/veracode-finds-ai-generated-code-security-has-barely-improved-since-last-year. 8sync News chỉ tóm tắt và dẫn link; bản quyền nội dung thuộc tác giả và nguồn gốc.