Anthropic just opened a new front.

On Thursday, Anthropic announced that Claude Security, its defensive cybersecurity product, is now available in public beta to Claude Enterprise customers worldwide. SiliconANGLE, ZDNET, The New Stack, and Inc. reported the launch within hours of Anthropic's own blog post. The tool scans codebases for vulnerabilities, validates findings to suppress false positives, and generates patch workflows for security teams to review.

Access for Claude Team and Max-tier customers is coming soon, per Anthropic's product page. Admins enable the feature in the admin console. The product graduated from a private research preview that started in February.

Why this lands today, not last week.

Anthropic shipped Claude Security on the same day the White House finalized the carveout that lets federal agencies bypass the Pentagon's supply-chain ban on Mythos. We covered that yesterday. The pattern is now visible. Anthropic is running a coordinated distribution sprint into every category that pays for security and reliability: federal agencies, creative software through the Adobe and Blender connectors, and now enterprise security teams.

The strategic logic is clean. Defensive cyber is the highest-margin enterprise SaaS category outside of databases. Crowdstrike trades at roughly 17x sales. Wiz sold to Google for $32 billion. If Anthropic can credibly position Claude Security as a code-scanning workhorse with autopatch, it converts model spend into seat-based licensing on top of existing Claude Enterprise contracts. That is exactly the revenue mix Anthropic needs heading into an October IPO that bankers are reportedly targeting at $400 to $500 billion.

The competitive read.

OpenAI does not have a defensive cyber product. Microsoft does, but it is locked inside the Defender stack and tied to Windows. Google has Mandiant but no Gemini-native scanner that touches the codebase. Anthropic just shipped one that runs on Claude Enterprise with admin-console toggles and patch generation. That is a clean shot at the part of the security market where AI-assisted code review actually has to work, not just look good in a demo.

There is a wrinkle. The same model that finds the vulnerability writes the patch. If the patch is wrong, you have a self-confirming bias loop running inside production code. Anthropic appears to have anticipated this. The product validates findings before surfacing them and routes patches through a human review step rather than auto-deploying. That is the right call. It is also the only call that gets it past enterprise security review.

What to watch.

Three things over the next 60 days. First, whether Microsoft accelerates a competing product inside Defender or Copilot, since Anthropic just attacked one of Redmond's highest-margin attach points. Second, whether the Pentagon carveout for Mythos extends to Claude Security inside classified environments, since defensive cyber is the most natural federal use case. Third, whether the IPO targeting reported by The Information moves up given how aggressively Anthropic is now monetizing every adjacent surface. The two-front sprint into creative tools and enterprise security is what a company does when the bankers are sharpening the prospectus.

First reported by SiliconANGLE, ZDNET, and The New Stack. Anthropic blog post on claude.com confirms availability and pricing tiers.