Semgrep Unveils Multimodal System for Enhanced Code Security

On March 19, 2026, Semgrep announced the launch of Semgrep Multimodal, a new system that integrates AI reasoning with rule-based analysis to improve detection, triage, and remediation of code vulnerabilities. This system reportedly identifies up to 8 times more true positives while reducing false positives by 50% compared to traditional foundation models. Semgrep Multimodal has already uncovered dozens of zero-day vulnerabilities at customer sites. The system is built on Semgrep Workflows, which allows security teams to automate their security processes across various environments. The launch addresses the growing gap between AI-generated code and existing security practices, which struggle to keep pace with the volume of pull requests. Traditional SAST tools excel at identifying known vulnerabilities but often miss complex business logic flaws, which Semgrep Multimodal aims to address. The new system is designed to enhance the reliability of security measures across organizations, leveraging both deterministic tools and AI. As underlying models improve, the performance of Semgrep Multimodal is expected to enhance automatically.

Key Points: • Semgrep Multimodal combines AI reasoning with rule-based analysis for improved vulnerability detection. • The system finds 8x more true positives and reduces noise by 50% compared to traditional models. • It has already discovered dozens of zero-day vulnerabilities at customer sites.