SHELFCRITTER AI Digest

PRODUCT

Claude Security

producttopic-noteanthropicsecurity

Overview

Claude Security is Anthropic’s repository-scale code-vulnerability scanning product, distinct from Claude Code Security. Launched in public beta on April 30, 2026, it is powered by Claude Opus 4.7 and positioned as a CISO-and AppSec-team product for scanning entire repositories, directories, or branches with reasoning over large codebases and complex dependency chains. The product targets enterprise CISO and AppSec orgs rather than individual developers, marking Anthropic’s entry into the security-SaaS market.

Timeline

Key Developments

  1. CISO-Tier Product Positioning: Claude Security targets CISOs and AppSec teams rather than individual developers, positioning vulnerability scanning as an enterprise security platform capability rather than a developer tool. Contrasts with Claude Code Security, which is integrated into Claude Code for developers.

  2. Codebase-Scale Reasoning: The product emphasizes reasoning over large codebases and complex dependency chains as its differentiator vs traditional per-file static analysis, leveraging Claude Opus 4.7‘s long context (1M tokens) and multi-step reasoning for security analysis.

  3. Three-Front Product Week: Claude Security GA is one of three Anthropic announcements in the week of April 28–May 3: (1) Claude Security on April 30, (2) Claude for Creative Work with nine first-party connectors on April 28, (3) Claude Personal Guidance sycophancy research on May 3. Signals the “agentic platform, not a model API” positioning across three distinct buyer audiences (CISO, designers, model-trust researchers).

  4. Commercial Enterprise Wedge: Launches as complement to Claude Code Security (developer-integrated, Enterprise-gated) and Claude Mythos Preview (Glasswing-gated federal infrastructure). Claude Security occupies the AppSec-buyer tier, expanding Anthropic’s security-product footprint across buyer tiers without cannibalizing existing tiers.

  • 2026-08-22-AI-Digest — Anthropic on 2026-08-21 deployed Claude Mythos 5 into Claude Security as an output-constrained deployment — the frontier Mythos-5 weights are now reachable only through the product’s structured scan interface (no prompt box, scan results only, per Anthropic “cannot be steered into writing exploits”), upgrading the model tier from the Claude Opus 4.7 baseline the beta shipped on. Distribution runs through five named SI channel partners — Accenture, BCG, Deloitte, Infosys, and PwC — for deployment and consulting into hospitals, utilities, and banks. An OEM path into third-party security vendors’ products is announced but not shipped (treat as roadmap, not deployed motion). A $35M open-source defense fund is attached to the launch as reputational counterweight to the constrained-surface productization. No pricing disclosed.

Key Developments

  1. Mythos 5 Deployed as Output-Constrained Scan-Only Surface With SI-Channel Distribution and $35M Open-Source Defense Fund (August 22, 2026): Anthropic on 2026-08-21 embedded Claude Mythos 5 — the same frontier model whose internal-only sibling “Model 2” was shelved in 2026-08-21-AI-Digest — inside Claude Security as an output-constrained deployment: the model is reachable only through the vuln-finding output surface, not a raw model API. Load-bearing framing to carry: separate three shipped-vs-announced things — (1) the scan-interface constraint is shipped and load-bearing; (2) the SI partner channel (Accenture / BCG / Deloitte / Infosys / PwC) is shipped as deployment and consulting, not co-selling of API access; (3) the OEM path into third-party security vendors is announced but not shipped — do not conflate the three. Do NOT read this as extending the 2026-08-21-AI-Digest “shelving” pattern — Mythos 5 is expanding access under a narrowed surface. Structural read: this is the first shipped frontier-lab instance of “output-constrained deployment” at production scale — the middle path the safety-tier motion of the week (Astra pause / GLM 5.3 weights delay / “Model 2” shelving) did not have: ship the capability, constrain the interaction surface. Whether other labs adopt it depends on whether an SI channel can actually monetize a model that customers can’t call directly. The $35M open-source defense fund is worth noting separately from the product story — reputational counterweight to a “we’re keeping the most capable model inside a walled product” posture; watch whether the grant list, when it publishes, resources defensive-tooling projects or reads as a PR line item. 30 / 60 / 90-day watch: whether the OEM-into-security-vendors path actually lands (productization test); whether OpenAI or DeepMind ship a comparable output-constrained surface on their own frontier tier (industry-motion test); whether the SI-channel arrangement produces disclosed customer wins with dollar figures inside the CISO buying centre (enterprise-monetization test).

See also: Claude Code Security, Claude Opus 4.7, Claude Mythos 5, Anthropic, MOC - Agent Security, MOC - Major Companies