Anthropic Launches Claude Code Security in Limited Preview to Scan Codebases for Hidden Vulnerabilities.

Anthropic has launched Claude Code Security, a new AI-powered tool that scans codebases for security vulnerabilities and suggests targeted patches for human review. The offering entered limited research preview today for Enterprise and Team customers, with expedited access available to open-source maintainers.

The tool aims to identify security issues that traditional static analysis and security scanning tools often miss. According to Anthropic's announcement, Claude Code Security represents a step toward raising the baseline security standard across the software industry, as AI-driven code analysis becomes increasingly capable of finding long-hidden bugs.

Claude Code Security analyzes entire codebases to identify potential vulnerabilities, then generates specific software patches for developers to review and apply. The human-in-the-loop approach ensures that security teams maintain control over which fixes get implemented, while benefiting from AI's ability to spot subtle issues that might escape manual review or rule-based scanning.

Anthropod positions the tool as complementary to existing security workflows rather than a replacement. Traditional tools excel at catching known vulnerability patterns and common misconfigurations, while AI models can identify novel security issues that don't match existing signatures.

The limited research preview is now open to customers on Claude's Enterprise and Team plans. Open-source project maintainers can apply for expedited access through a dedicated application process, recognizing the outsized impact that securing widely-used open-source dependencies can have on the broader software ecosystem.

Anthropoc did not specify timeline for general availability or pricing details beyond the existing Enterprise and Team tier requirements.

The launch follows Anthropic's recent Project Glasswing announcement, which uses Claude models to secure critical software infrastructure. Code security represents a natural expansion of that effort, bringing AI-powered vulnerability detection to a broader range of codebases.

Other AI companies have made similar moves into code security. GitHub Copilot added security vulnerability detection last year, while several startups have launched tools that use large language models to identify and fix security issues in production code.

The effectiveness of AI models at finding security vulnerabilities has improved dramatically over the past year. Models can now understand complex code patterns, trace data flows across multiple files, and identify subtle logic errors that create security holes. Anthropic expects a significant portion of the world's code will be scanned by AI in the near future as these capabilities mature.

For developers building with AI coding assistants like Cursor or Windsurf, Claude Code Security could provide an additional layer of security review before code reaches production. The tool may prove particularly valuable for small teams that lack dedicated security engineers but still need to maintain secure codebases.

Enterprise and Team customers can access Claude Code Security through their existing Claude accounts starting today. Open-source maintainers can apply for access through the link provided in Anthropic's announcement.