Anthropic launches Claude Security in public beta for enterprise customers

Anthropic has launched Claude Security in public beta for Claude Enterprise customers, giving organizations a built-in way to scan codebases for vulnerabilities and generate proposed fixes.

The tool uses Claude Opus 4.7 and is designed for defensive security work. It can review repositories, reason across files and modules, validate findings, and suggest targeted patches that developers can review before applying.

The launch gives security teams a direct route to AI-assisted vulnerability analysis without building custom agents, wiring API integrations, or creating separate internal tooling.

Claude Security moves from preview to public beta

Claude Security was previously known as Claude Code Security. Anthropic introduced it as a limited research preview in February 2026 for Enterprise and Team customers.

Since then, hundreds of organizations have tested it on production codebases. Anthropic says that feedback shaped the public beta version now available to all Claude Enterprise customers.

Access for Claude Team and Max customers is expected later, but Anthropic has not given a specific rollout date for those plans.

At a glance

Category	Details
Product	Claude Security
Company	Anthropic
Status	Public beta
Available to	Claude Enterprise customers
Model used	Claude Opus 4.7
Main function	Codebase vulnerability scanning and proposed fixes
Previous name	Claude Code Security
Setup requirement	No custom API integration or agent build required

How Claude Security works

Enterprise users can start from the Claude.ai sidebar or go directly to Claude Security. They select a repository, choose the scan scope, and launch the review.

The scan can cover a full repository, a specific branch, or a targeted directory. That gives teams a way to focus on sensitive code paths, active development areas, or high-risk modules.

Anthropic says Claude Security does more than search for known vulnerability patterns. It traces data flows, reads source code, and studies how components interact across the codebase.

What Claude Security reports

• Detailed vulnerability explanations
• Confidence ratings for each finding
• Severity and likely impact
• Reproduction guidance
• Suggested patch instructions
• Exportable findings for tracking and audits
• Webhook notifications for workflow tools

Validation aims to reduce false positives

False positives remain one of the biggest problems in security scanning. A tool that reports too many weak findings can slow down both security teams and engineering teams.

Anthropic says Claude Security uses a multi-stage validation pipeline before a finding reaches an analyst. The goal is to filter weak results and raise the quality of alerts.

Each finding includes a confidence rating, which helps teams decide what to investigate first. This matters because security teams often work through large backlogs with limited time.

New features added for the public beta

Feature	Why it matters
Scheduled scans	Teams can run recurring security reviews instead of relying on one-time checks.
Directory-level targeting	Security teams can focus on specific modules or high-risk paths.
Branch targeting	Teams can scan active development work before changes move further downstream.
CSV and Markdown exports	Findings can move into existing reporting, audit, and tracking systems.
Webhook notifications	Results can flow into tools such as Slack, Jira, and other internal systems.
Documented dismissals	Teams can carry triage decisions forward with context for future reviewers.

Why Anthropic is positioning this for defenders

Anthropic says AI is changing the speed of vulnerability discovery and exploitation. As models improve, attackers may be able to find and weaponize flaws faster.

Claude Security is Anthropic’s attempt to put similar code-analysis power into defensive workflows. The company is positioning it as a way for organizations to review more code without adding more manual triage.

The timing also follows Anthropic’s work on Project Glasswing, a separate effort focused on frontier cybersecurity capabilities for a limited set of partners.

Claude Opus 4.7 powers the product

Claude Security runs on Claude Opus 4.7, Anthropic’s latest generally available Opus model. Anthropic says the model improves on Opus 4.6 in advanced software engineering and long-running coding tasks.

For security work, Anthropic says Opus 4.7 can help find and patch complex, context-dependent vulnerabilities that simple pattern matching may miss.

The model also includes cyber safeguards designed to detect and block requests linked to prohibited or high-risk cybersecurity uses. Anthropic says legitimate security professionals can apply for its Cyber Verification Program when their work may trigger those safeguards.

Partners will bring Opus 4.7 into security tools

Claude Security is not the only route Anthropic is using for enterprise security adoption. The company also says several security vendors are embedding Opus 4.7 capabilities into their own platforms.

The partner list includes CrowdStrike, Microsoft Security, Palo Alto Networks, SentinelOne, TrendAI, and Wiz.

Anthropic also named Accenture, BCG, Deloitte, Infosys, and PwC as services partners helping organizations deploy Claude-integrated security solutions.

What enterprise teams can use it for

• Vulnerability discovery in production repositories
• Secure code review before release
• Prioritizing high-confidence findings
• Generating patch instructions for developers
• Reducing manual back-and-forth between security and engineering teams
• Exporting findings for audit and compliance workflows
• Adding recurring scans to vulnerability management programs

Why this matters for developers

Security findings often fail to move quickly because developers need context. A ticket may describe a vulnerability, but it may not explain the affected code path, impact, or exact fix.

Claude Security tries to shorten that gap by producing findings and patch guidance in the same workflow. Developers can then open the issue in Claude Code on the Web and work through the fix in context.

That could help teams reduce the time between scan and remediation, especially when the issue requires understanding business logic across multiple files.

Limits enterprises should keep in mind

Claude Security should not replace human review, secure design, dependency management, penetration testing, or traditional application security testing.

AI-assisted scanning can improve coverage, but teams still need clear ownership, review controls, patch validation, and release safeguards.

The strongest use case is likely as part of a layered application security program, where Claude Security helps surface deeper findings and reduce triage work.

FAQ