Anthropic wants to help developers catch security vulnerabilities before cybercriminals exploit them. Today, the AI company rolled out OSS Scanner, a free service that uses its most powerful models—including Claude Mythos—to run automated code audits across open-source projects.
As hackers increasingly rely on AI to discover zero-day bugs and automate cyberattacks, Anthropic is giving developers access to the same high-level scanning tools. Open-source maintainers can sign up by opening a pull request on GitHub and providing a Dockerfile, allowing offline AI agents to clone and audit their codebase in an isolated environment without internet access.
Skipping human review means dealing with false alarms
To keep security checks fast and frequent, Anthropic decided to skip human review entirely. Every report generated by OSS Scanner lands directly in the maintainer’s inbox without an analyst checking it first.
The obvious downside here is false alarms. AI models can easily misread code and flag non-existent bugs, an issue that developers at Google and Linux creator Linus Torvalds have complained about during recent waves of automated reports. Knowing this, Anthropic is waiving strict 90-day public disclosure deadlines for these reports unless a human manually verifies the flaw down the line.
Building automated defenses for critical software
OSS Scanner builds on lessons learned during Anthropic’s internal Project Glasswing, which uncovered over 29,000 candidate bugs across major software repositories. That effort produced more than 6,000 direct reports to developers and nearly 600 official security advisories.
This new tool arrives alongside Anthropic’s broader Critical Infrastructure Defense Program, aimed at safeguarding essential software networks. While AI scanners recently helped uncover severe issues like the “Copy Fail” bug that hit almost every Linux distribution, Anthropic expects AI progress over the next two years to heavily favor cyber defenders by making automated code patching and real-time security auditing standard practice.
The post Anthropic Is Offering Free AI Bug Hunting for Open-Source Projects appeared first on Android Headlines.
