Anthropic Opens Claude Security Public Beta to Hunt Vulnerabilities Across Whole Codebases
Powered by Opus 4.7 and the Mythos-fueled Project Glasswing, Claude Security reasons over entire codebases like a human researcher — tracing data flows instead of matching patterns.
Anthropic moved Claude Security out of research preview and into public beta on April 30, putting an autonomous vulnerability hunter directly in the hands of Claude Enterprise customers. Accessed from a sidebar at claude.ai/security, the tool scans codebases to find software flaws and proposes patches, with Anthropic positioning it as an answer to a wave of AI-powered exploit campaigns hitting enterprises.
The pitch is that Claude Security does not look for known patterns. Built on the company’s newly generally available Opus 4.7 model and its Project Glasswing pipeline (which leans on the Mythos model for vulnerability discovery), the system traces data flows, reads source code, and examines interactions between components much the way a human security researcher would. During the research preview, Anthropic says hundreds of organizations used it to surface and patch issues that legacy scanners had missed for years.
To tame false positives, every finding runs through a multi-stage validation pipeline before it reaches an analyst, complete with a confidence rating and a written explanation of reasoning, exploit likelihood, triage factors, and fix effectiveness. Engineers can apply patches immediately inside a Claude Code session, removing the back-and-forth between security and dev teams. The beta also adds scheduled scans, the ability to dismiss findings with documented reasons, and CSV plus Markdown exports for existing audit workflows.
Anthropic is launching with a heavyweight partner roster. CrowdStrike, Palo Alto Networks, SentinelOne, Trend Micro’s TrendAI, and Wiz are integrating Opus 4.7 into their cybersecurity platforms, while Accenture, BCG, Deloitte, Infosys, and PwC are building Claude-powered services for vulnerability management, secure code review, and incident response. Access for Claude Team and Max plan customers is expected to follow soon, as Anthropic races to convert frontier-model gains into recurring enterprise revenue.
