Anthropic Opens Claude Security to Enterprises in Public Beta for Vulnerability Scanning

Anthropic on April 30, 2026 moved Claude Security out of private trials and into a global public beta available to Claude Enterprise customers, according to SiliconANGLE, SecurityWeek, and Infosecurity Magazine. Running on the company's most capable generally available model, Claude Opus 4.7, the product reads source code across files, traces data flows, and reasons through logic-level flaws rather than only matching known vulnerable patterns. Anthropic told reviewers that, in earlier evaluations, expert validators agreed with Claude's severity assessment exactly in 89% of 198 manually reviewed vulnerability reports — the public-beta number Anthropic is leading with on positioning vs. traditional static-analysis tools.

The April 30 release adds the operational glue enterprises had been asking for. Per BankInfoSecurity and SiliconANGLE, the beta now ships scheduled scans, documented dismissals so triage decisions are auditable, and CSV or Markdown exports that drop directly into ticketing systems. Customers can also push findings to internal tools via per-project webhooks, and scans can target a full repository, a specific directory, or an individual branch. A multi-stage validation step runs each candidate finding through additional model passes to drive down false positives — the failure mode that has historically gated AI security tooling from production use.

Two go-to-market choices stand out. First, Claude Security is included for existing Claude Enterprise users at no extra integration cost, sidestepping the procurement friction of standalone AppSec products. Second, Anthropic is gating high-risk capabilities behind a separate Cyber Verification Program: users who haven't verified themselves as cybersecurity professionals see guardrails on requests that look offensively oriented, while verified researchers and defenders unlock the full surface. The launch positions Claude Security against incumbents like Snyk and GitHub Advanced Security and against Anthropic's still-restricted internal model, Mythos — the latter remaining behind closed access while Opus 4.7 takes the public-facing role on enterprise code defense.