AI
Anthropic offers free, opt-in security scans to open-source projects
Anthropic has opened a free service that scans open-source software for security holes, which are called vulnerabilities. Eligible projects get periodic scans from Anthropic's strongest models, and maintainers enroll by sending a pull request (a proposed code change) to anthropics/oss-scanner.
In Anthropic's own six-month trial it found over 29,000 candidate vulnerabilities. Penetration testers (people paid to break into things) checked 97 of the most serious across 48 projects; 85 (88%) were worth reporting, with one false alarm. Caveat: the reports are fully model-written with no human sorting, so some may be wrong.
On your desk: you probably will not enroll a project, but it shows how good AI code review is getting, which is the kind of help your Claude Code builds will inherit. Low priority.
Read the original at Anthropic: opt-in vulnerability finding service for open source →