AI-Generated Code Introduces 4.4x More Vulnerabilities, 98% of Vibe-Coded Apps Flawed

New research quantifies that AI-generated code has 4.4x more security vulnerabilities than human-written code, with 98% of vibe-coded applications containing flaws. This challenges the assumption that AI code is safe for production and reinforces the need for shift-left security and AI-specific tooling. The study provides concrete data to justify investment in AI security controls. GitLab 19.3 responds with Dedicated AI Gateway, Secrets Manager, and Bulk SAST Remediation to secure agentic development and address the vulnerability bottleneck.