Anthropic opens Claude Security for code vulnerability scans

Anthropic has opened Claude Security in public beta for Claude Enterprise customers, turning Claude into a codebase security scanner rather than only a coding assistant. The product page says Claude scans repositories for vulnerabilities, validates findings and suggests patches that teams can review before applying.

The companion tutorial says the feature is built into Claude.ai and currently supports repositories hosted on GitHub.com. Enterprise admins need the Anthropic GitHub App installed, access to the target repositories, and Extra Usage enabled because scans use consumption billing. Anthropic says premium-seat users can run scans on Claude Mythos 5.

The workflow is aimed at teams that do not have enough security reviewer time for every code change. Users choose a repository, optionally scope the scan to a directory or branch, then review findings with vulnerability type, severity, file, line and suggested fix. A remediation button can open a Claude Code on the Web session focused on the specific issue.

The useful signal is that Anthropic is packaging frontier models into a concrete enterprise security workflow. The caveat is the usual one: patches and findings still need human review before production use.