According to Anthropic, the scanner draws on the company's experience using Claude to uncover vulnerabilities during Project Glasswing. Projects that enroll will receive comprehensive, recurring security scans performed by Anthropic's most capable models, at no cost to the maintainers.

Key points:

• OSS Scanner is an opt-in service, meaning projects must actively join to be scanned.
• Scans are periodic and performed by Anthropic's strongest models.
• The service is offered free of charge to participating open-source projects.
• The tool was informed by Anthropic's prior work using Claude to find vulnerabilities in Project Glasswing.

Source: The Hacker News

Would you trust AI-driven scanning to catch vulnerabilities in the open-source projects you depend on, or would you still want a human review in the loop?