sglang — Vulnerabilities & Security Advisories 18

This page documents security vulnerabilities affecting the sglang large language model inference engine, categorized under software weakness types such as input validation errors and resource management flaws. The content aggregates known vulnerabilities from public sources, including Common Vulnerabilities and Exposures (CVE) databases and vendor security advisories, covering incidents reported from the initial release of sglang through the present day. Readers can utilize this resource to track a specific vendor's security posture by monitoring their advisory history, gain a deeper understanding of common weakness classes prevalent in high-performance LLM serving frameworks, or look up the complete vulnerability history of the sglang product to assess risk exposure. By consolidating these data points, the page provides a centralized view for developers, security engineers, and system administrators to evaluate the integrity and stability of their deployments. The information presented is intended to support informed decision-making regarding patch management and configuration hardening without implying any specific liability or endorsement. All entries are sourced from publicly available disclosures to ensure transparency and accuracy, allowing users to verify details through original references where applicable. This structured approach helps in identifying patterns in flaw types and understanding the evolution of security issues within the sglang ecosystem over time.