Anthropic launches OSS Scanner for free AI security checks on open source

In short: Anthropic has launched OSS Scanner, a free opt-in service that uses its AI models to look for security problems in open-source software.

What happened

Anthropic announced a new service called OSS Scanner to help open-source projects find security vulnerabilities. A vulnerability is a weakness in software that could let someone break in, like a door that does not fully latch.

Open-source projects can choose to opt in, meaning they sign up on purpose to be scanned. Anthropic says it will run thorough, periodic scans using its “strongest models,” including one it calls Claude Mythos, at no cost.

There is an important catch. Anthropic says the reports are fully generated by AI and do not include human review or “triage,” which is the step where a person checks what matters most. That means some reports could be wrong or not actually a real problem.

Why it matters

A lot of the software people rely on, including parts of the internet and many apps, depends on open-source code maintained by small teams or volunteers. Faster warnings about possible bugs could help those teams fix issues sooner. At the same time, AI-generated reports can add extra work if they include false alarms, and some projects have already struggled with large waves of low-quality bug reports.

Source: The Verge AI