Claude Security checks code at each change and marks repairs

Anthropic released a new version of Claude Security in public beta for Claude Enterprise customers. Anthropic reports that thousands of organizations used it since April and found more than one million vulnerabilities. The tool checks a repository (where a team keeps its code) each time the code changes. Anthropic gives $25,000 of free credits each month for two months.

What is new

Claude does a deep first check. Then it scans each change at a time set by its risk and complexity. A project can hold many repositories, and Claude can find vulnerabilities between services. Each project has a knowledge base where a team writes about its controls. Claude also makes a threat model from the code, and the team can edit it.

Control and access

• A new API (a connection for other software) lets teams read vulnerabilities, resolve them or open them again.
• Webhooks (automatic messages) tell other systems about a new vulnerability.
• Role-based controls set who can view projects and work on vulnerabilities.
• An MCP connector (a standard connection between tools) works in Claude.ai and Claude Code.

What is not known

Anthropic gave no test results on how many vulnerabilities are correct. The credits apply to use through December 31, 2026. The source text gives no cost after the credits.

Insight: Anthropic wants to hold the record of vulnerabilities

On 15 October, GitHub made AI Scan available for pull requests, but its results cannot stop a merge. Anthropic adds an API, webhooks and a record of repairs to Claude Security. This reads as a move to hold the record of vulnerabilities, because that tool is the one teams use all the time. Select which tool your team wants.