After a record month of crypto exploits, Anthropic opened Claude Security to public beta. The tool scans codebases for high-severity vulnerabilities including injection flaws, authentication bypasses, and logic errors that pattern-matching scanners typically miss.

Unlike traditional tools, it runs an adversarial verification pass on its own results before surfacing them, reducing false positives without requiring a rule update.

Every suggested fix requires human review before deployment. Available now for Claude Enterprise, Team and Max coming soon.