Anthropic unveils Claude Code security to detect hidden vulnerabilities in software codebases

Anthropic has launched Claude Code Security, an AI-powered tool that scans codebases for complex security flaws, aiming to reduce vulnerability backlogs for developers and security teams.

AI research firm Anthropic has introduced Claude Code Security, a new feature designed to help developers and security teams identify and fix vulnerabilities across large software codebases. The tool is currently available in a limited research preview.

Unlike traditional rule-based static analysis tools, Claude Code Security uses artificial intelligence to understand how code actually behaves. This allows it to uncover deeper, context-dependent issues such as logic flaws, insecure access controls, and risky data flows that are often missed by conventional scanners.

Each potential vulnerability detected by the system goes through a multi-stage verification process. Claude rechecks its own findings to reduce false positives and assigns severity levels along with confidence scores. The validated results are then displayed in a dedicated dashboard, where teams can review recommended fixes.

Importantly, no changes are applied automatically. Developers retain full control, reviewing and approving all remediation steps before any fixes are implemented. This human-in-the-loop approach is intended to balance automation with accountability.

The launch comes as security teams worldwide struggle with growing vulnerability backlogs. Anthropic says the tool builds on more than a year of internal cybersecurity research, including testing by its Frontier Red Team through hackathons, real-world exercises, and collaborations with institutions such as the Pacific Northwest National Laboratory.

Powered by the latest Claude Opus 4.6 model, the system has reportedly helped identify more than 500 previously undetected vulnerabilities in open-source projects, some of which had gone unnoticed for decades.

Anthropic positions Claude Code Security as a defensive tool in an era where AI is increasingly used by attackers as well. The company envisions a future where intelligent systems continuously scan global codebases, predicting and preventing threats before they can be exploited.

Enterprise and team customers can now request early access, while open-source maintainers are invited to join a fast-tracked preview program.