Anthropic Launches Claude Security Public Beta

Anthropic on Thursday launched Claude Security into public beta for all Claude Enterprise customers, offering an AI-powered tool to scan codebases for vulnerabilities and generate targeted fixes amid rising AI-driven cyber threats.

The tool, powered by Anthropic's Claude Opus 4.7 model released in mid-April, builds on a limited research preview that began in late February under the name Claude Code Security. Hundreds of organizations tested it during that phase, providing feedback that led to enhancements like improved detection accuracy, reduced false positives and confidence scoring for findings. Administrators can now enable the feature through the Claude admin console, with access for Team and Max customers to follow soon. Users access it via the sidebar on Claude.ai or directly at claude.ai/security.

Unlike traditional scanners reliant on pattern matching, Claude Security analyzes code like a human researcher. It examines source code across files and modules, traces data flows and assesses component interactions to identify complex, context-dependent issues. For each vulnerability, it delivers detailed reports including confidence levels, severity ratings, reproduction...