Anthropic has launched Claude Security in public beta for Enterprise customers, an AI-powered tool that scans entire codebases for vulnerabilities and generates targeted patches. Powered by Claude Opus 4.7, it traces data flows and component interactions rather than relying on pattern matching, surfacing issues that traditional static analysis tools miss. Each finding includes confidence ratings, severity levels, reproduction steps, and recommended fixes. New features include scheduled scans, dismissal with documented reasoning, and CSV/Markdown export. Alongside the launch, Anthropic introduced Project Glasswing for securing critical infrastructure using the restricted Mythos model, and announced integrations with CrowdStrike, Palo Alto Networks, SentinelOne, and major consulting firms.
Table of contents
From Research Preview to Public BetaHow it WorksBuilt for How Security Teams Actually WorkPart of a Broader Security PushStrong Partner Ecosystem From Day OneWhat This Means for DevSecOpsSort: