Chainguard Launches AI Tool for Application Vulnerability Discovery
Chainguard has introduced Chainguard Harness, a cybersecurity tool that uses frontier AI models to find and fix exploitable vulnerabilities in application code without requiring direct access to the AI models.
Seattle, WA – Chainguard has launched Chainguard Harness, a cybersecurity tool designed to identify and fix exploitable vulnerabilities in application code. The tool leverages advanced AI models, including Anthropic's Claude Mythos 5.1, to analyze an organization's application by building it in an isolated microVM environment. Harness then probes the live, running instance for vulnerabilities and provides ready-to-merge fixes.
A key benefit of Harness is its ability to provide organizations with the offensive security capabilities of frontier AI without the need for them to manage their own complex AI models. Unlike traditional scanners that analyze static code and may produce false positives or miss runtime-specific issues, Chainguard Harness tests the application in its operational state. This approach aims to uncover logic flaws and injection paths that only appear during execution, thereby reducing noise and focusing on proven vulnerabilities.
"Every time we've pointed Harness at code, it has found serious, exploitable vulnerabilities, including in software that's been scanned many times over," said Dan Lorenc, CEO and Co-founder of Chainguard. "Harness offers security teams a way to get proven vulnerabilities and fixes their engineers can ship."
Chainguard Harness is available through the Claude Marketplace by Anthropic. The tool connects to a team's code repository via GitHub, builds the application, and boots it in an isolated environment. The AI then hunts for vulnerabilities within the live application. All findings are reproduced and verified before fixes are provided, with the corrections tailored to the specific codebase.