Hadrian Launches OpenHack for AI Vulnerability Research

Hadrian has open-sourced OpenHack, a file-based workflow for AI-assisted source-code vulnerability research, announced on May 20, 2026. This MIT-licensed tool integrates with Claude Code, Codex, and Cursor, enabling automated vulnerability assessments. OpenHack employs a scenario-first review model and independent triage to enhance accuracy and reduce false findings. The methodology was previously used to identify critical vulnerabilities in open-source software utilized by Dutch government agencies. The system maintains a durable state in plain files and includes a registry of expert families aligned with OWASP and MITRE standards. OpenHack aims to democratize advanced vulnerability research methodologies, making them accessible to a broader audience. The tool's design focuses on reproducibility and auditability, addressing common issues in model-driven code reviews.

Key Points: • OpenHack is an open-source tool for AI-driven source-code vulnerability research. • The tool integrates with popular coding harnesses like Claude Code, Codex, and Cursor. • Hadrian's methodology previously uncovered critical vulnerabilities in software for Dutch government agencies.