Anthropic launches free AI-driven security scanner for open-source code

Anthropic has launched the OSS Scanner, a free initiative designed to protect open-source software projects by using advanced AI models to identify security vulnerabilities. The service provides periodic, automated scans of codebases to offer early warnings about potential exploits, aiming to support small teams and volunteers who maintain critical infrastructure.

The scanning process is powered by Anthropic’s most capable frontier models, including Claude Mythos. To ensure high frequency and speed, the reports are entirely model-generated without human triage. Anthropic noted a tradeoff between speed and accuracy, warning that maintainers may encounter false positives or invalid reports that require manual filtering and testing before implementation.

In addition to general open-source projects, the company is dedicating a specific program to protecting critical operational technology, such as electrical grids, water supplies, and transportation systems. This initiative was inspired by Google’s OSS-Fuzz service.