Anthropic puts Claude Security in beta

Security tooling is the thing here — not just another chatbot feature. Anthropic has pushed Claude Security into public beta, which means its AI vulnerability scanner is now available to Claude Enterprise customers instead of staying tucked inside a limited preview. The pitch is simple: point Claude at a codebase, let it reason through how data actually moves, then get back prioritized bugs and patch suggestions a human can approve. Anthropic first showed the capability as “Claude Code Security” in a research preview on February 20, 2026, and the public-beta move turns that experiment into a product. (claude.com) ### What is Claude Security, exactly? It’s an AI-assisted code security product built around full-codebase analysis. Anthropic says it scans repositories for vulnerabilities, validates whether the findings are real, and recommends patches for human review and approval. The company frames the advantage as context — Claude can trace data flows and understand how pieces of a codebase connect, instead of relying only on pattern matching or static rules. (clau([claude.com)What changed in this beta? The big change is availability. Anthropic’s product page now labels Claude Security “available in public beta” for Claude Enterprise. Coverage also expanded beyond the earlier preview framing: the current beta includes scheduled scans and targeted scans, which makes it feel less like a one-off demo and more like something a security team could fit into normal engineering workflows. (claude.com) Because the real fight in software security is not finding one scary bug in a demo repo. It’s catching issues repeatedly, inside the release cycle, before they pile up. Scheduled scans mean teams can run recurring checks across the same codebase. Targeted scans mean they can zero in on specific changes or areas of concern. Basically, Anthropic is trying to move security review closer to CI/CD and away from the old(claude.com)hat is a practical shift, not just a branding one. (msn.com) ### What problem is Anthropic trying to solve? Traditional scanners are good at known patterns, but they often flood teams with noisy alerts or miss bugs that depend on application context. Anthropic’s pitch is that a frontier model can reason more like a skilled reviewer — follow inputs, permissions, and execution paths across files — then surface fewer, higher-confidence issues. The c(msn.com)than automatic deployment. (claude.com) ### Is this separate from Claude Code? Not really. The lineage runs straight through Claude Code. Anthropic introduced automated security reviews in Claude Code back in August 2025 with a `/security-review` command and GitHub Actions integration. The February 2026 preview then packaged that idea into a more formal security capability. Claude Security looks like the enterprise productization of that same direction — less “ask the assistant for help,” more “give the security team a dedicated workflow.” (claude.com) ### Who is this for? Large development teams, basically. Anthropic is gating the beta behind Claude Enterprise, not the general consumer product. That tells you the target buyer is an engineering or security organization that wants repository-wide analysis, repeatable scans, and some auditability around suggested fixes. This is enterprise software territory — fewer vibes, more workflow integration. (claude.com)ybersecurity has become one of the clearest enterprise use cases for advanced models. Anthropic has spent the past few months talking about frontier models as tools for defenders, including its broader push to make high-end cyber capabilities available to security teams. Claude Security is the commercial expression of that idea: if these models can find and help fix real flaws, Anthropic wants to sell that capability directly. (anthropic.com) ### Bottom line? Claude Security in public beta is Anthropic trying to turn AI code reasoning into a real security product — one that sits inside enterprise development loops instead of outside them. If it cuts false positives and saves AppSec time, it matters. If it just adds another dashboard, it won’t. (claude.com)