The latest version of Claude Security is in public beta for Claude Enterprise customers, Anthropic announced on October 21, 2026. The announcement mentions only Claude Enterprise customers. Organizations get $25,000 in free usage credits each month for two months, for usage through December 31, 2026. The new version keeps scanning as code changes and tracks whether findings get fixed. Tracking a fix means Claude Security checks whether a fix has shipped and marks the finding fixed; it does not fix vulnerabilities on its own. We have not tested it, so everything below follows Anthropic's own description.

One-time review vs continuous monitoring

Anthropic frames the change simply: a scan captures one moment, and code changes every day. Until now, someone had to start a scan. Since its public beta began in April, Anthropic says, thousands of organizations have used Claude Security to find over a million vulnerabilities, but each result reflected the code at the time of that scan.

In workflow terms, the difference looks like this:

  • Who starts it. A one-time review waits for someone to run it. Continuous monitoring runs on its own after setup.
  • How current the list is. A review's findings age as soon as new commits land. Continuous scanning aims to keep the findings list close to the current code.
  • What happens after a fix. With a one-time review, a person has to confirm the fix and close the item. Claude Security now checks open findings against changed code and marks them fixed.
  • What happens if it comes back. A closed item from a past review stays closed. Here, a finding reopens automatically if Claude Security finds the issue again.

Continuous scanning as code changes

Teams create a project, connect repositories through the Claude GitHub app (including GitHub Enterprise Server) or with an access token from GitLab and other Git providers, and choose the branch and scope for each. Claude runs a deep baseline scan, then scans changes regularly based on their risk and complexity.

Checks across repositories

A project can hold multiple repositories, so Claude can catch issues that sit between services. Anthropic's example is a backend that trusts input a client was supposed to validate.

Fix-status tracking

Most fixes happen outside the tool that found the problem. Claude Security checks open findings against the code that changed and marks them fixed when the fix lands, whether the team fixed it in Claude Code or elsewhere. If the issue shows up again later, the finding reopens automatically. This confirms whether a fix shipped; it is not automatic remediation.

Team context and threat models

Each project has a knowledge base where teams can describe how code is deployed, which security policies apply, and anything a new security engineer would need to know. Claude Security also generates a threat model from the code that the team can review and edit. Later scans use that threat model to decide what is worth flagging, and added context helps skip risks the environment already handles. Each finding comes with a severity rating judged against the threat model and an explanation of how an attacker could reach the vulnerable code, sometimes with an attack-path diagram.

Anthropic has used Claude on security work before, as in its attack surface reduction study across 25 open-source projects.

API, webhooks and MCP

The new API lets teams list and read projects and findings, resolve or reopen findings, and create and manage projects and their repositories. Users create a personal access token in Claude Security settings, and the token can only see and change what that user can. Webhooks can notify other systems when a new finding lands.

Claude Security is also an MCP connector in Claude.ai and Claude Code. Anthropic describes an engineer pulling up a finding, asking Claude about it, starting the fix and marking it resolved in the same session.

Access and models

New role-based access controls let admins decide who can view projects, triage findings and create projects. Admins enable Claude Security in the admin console. Anthropic says it runs on its full family of models, including Claude Mythos, with no Cyber Verification Program access required.

Continuous scanning watches your own code. It is a different problem from a compromised dependency, such as the credential stealer shipped in LiteLLM 1.107.0, and the announcement does not say how Claude Security handles that case.