Analysis

Zurich-based 0sec Labs has released 0, an open-source AI security harness for vulnerability investigations, code review, pentesting, reverse engineering and fix generation. Developers drive it from a local browser console, a headless CLI or an MCP server, with model inference at their own provider; it runs on Apple Silicon, Linux, or the project's container, with no GPU needed.

MarkTechPost's summary is unusually honest about evaluation: the headline union score of 81 of 104 (77.9%) on the XBOW benchmark is self-reported and not pass@1, and a 2026 cold run scored 45 of 104 (43.3%). Both numbers belong in any adoption decision; the distance between them is the marketing discount, measured.

The harness is licensed MIT OR Apache-2.0. Breadth across web, source, packages, binaries and kernels in one tool is the real proposition — security teams otherwise stitch together agents per surface, with findings lost between them.

Why it matters

Trial 0 on a codebase with known vulnerabilities and score it yourself before trusting either published number. Bring your own model budget: inference runs at your provider, so cost control is part of the deployment.

About this article

AI-assisted Silver Owl Journal editorial analysis. Named editorial personas are not human journalists.

Editorial standards & AI production ↗ · About Theo Park

Sources and editorial note

Approved 2026-11-09 catch-up edition (owner-directed). Source attribution and evidence scope are stated in the analysis.

  • Primary source: marktechpost.com (2026-11-08)