Overview
Claude Security is Anthropic's product line for enterprise-grade security scenarios, optimized for network defense, critical infrastructure protection, security research, and incident response. The core model is the Claude Mythos series—a specialized branch of Claude trained for cybersecurity capabilities, running parallel to the mainline Claude Opus/Sonnet series. The branch first appeared as Claude Mythos Preview in early 2026, the first public preview of the cybersecurity-specialized Claude line.
Three major milestones in 2026:
❶ 2026/4/7 Project Glasswing launched: Anthropic announced an investment of $100 million in model usage credits, partnering with leading security organizations (CrowdStrike, Microsoft, Palo Alto Networks, Trellix, etc.) to establish 'Project Glasswing', with the goal of protecting critical software security in the AI era. Why the name Glasswing (Glasswing Butterfly)? It symbolizes 'AI making security defense transparent'—as attackers use AI to accelerate attacks, defenders must also use AI to respond.
❷ 2026/6/9 Claude Mythos 5 officially released: Officially claimed to be the most capable frontier model for cybersecurity on the market, covering:
- Vulnerability detection and code audit: Identifies complex vulnerability patterns, with 40%+ higher accuracy than general models
- Threat intelligence analysis: Processes massive security logs, correlates events, and generates incident reports
- Malicious code analysis: Deobfuscation, anti-debugging code understanding, and behavioral analysis
- Red and blue team assistance: Supports red team penetration testing and blue team defense response
- Compliance audit: Automated mapping of SOC 2, ISO 27001, NIST, and other compliance requirements
❸ 2026/10/9 Free AI scanner for open source projects: Anthropic launched a free AI scanner for open source projects that periodically checks repositories, automatically flags and explains potential issues, and suggests patches. The company notes that reports are not human-reviewed and may be wrong, and maintainers of critical projects can opt in through GitHub.
Project Glasswing now covers approximately 150 organizations across 15 countries—including critical infrastructure protection, network defense research, and government agency collaborations. This is Anthropic's largest-scale deployment of its 'AI for the benefit of humanity' strategy in the security domain.
Product form: Not a standard API or subscription service, but a customized solution for enterprise customers—accessed through the Claude Enterprise commercial channel, with dedicated professional security consulting teams and exclusive access to the Mythos 5 model. Suitable for large enterprises, government agencies, and critical infrastructure operators (power/finance/healthcare/transportation).
Key Features
- Claude Mythos 5 Specialized Model: A Claude branch specifically trained for cybersecurity scenarios, the most capable frontier model for network security on the market
- Project Glasswing Ecosystem: $100 million in model credits + a joint defense network of 150 organizations across 15 countries
- Free AI Scanner for Open Source Projects: Periodically checks repositories, automatically flags and explains potential issues, and suggests patches; maintainers of critical projects can opt in through GitHub
- Vulnerability Detection and Code Audit: Identifies complex vulnerability patterns with 40%+ higher accuracy than general models
- Threat Intelligence Analysis: Processes massive security logs, correlates events, and automatically generates incident reports
- Malicious Code Analysis: Deobfuscation, anti-debugging code understanding, and behavioral chain analysis
- Red and Blue Team Assistance: Supports red team penetration testing and blue team defense response workflows
- Compliance Audit Automation: Automated mapping and gap analysis of SOC 2, ISO 27001, NIST, and other compliance requirements
- Critical Infrastructure Protection: Customized solutions for critical infrastructure in power, finance, healthcare, and transportation
Use Cases
- Automated incident response for enterprise SOC (Security Operations Center)
- AI-driven defense for critical infrastructure (power/finance/healthcare)
- Cybersecurity defense and intelligence analysis for government agencies
- AI underlying capabilities for security vendors (CrowdStrike / Palo Alto, etc.) products
- Red team penetration testing and blue team defense drill assistance
- Automation of compliance audits (SOC 2/ISO 27001/NIST)
- AI security research (model adversarial attacks, jailbreak defense, content security)
- Periodic security review of open source project repositories via the free AI scanner
Pros
- Claude Mythos 5 is the most capable frontier model for cybersecurity on the market
- Project Glasswing's $100 million investment demonstrates Anthropic's strategic commitment
- Scaled deployment validated across 150 organizations in 15 countries
- Covers full spectrum: vulnerability detection, threat intelligence, malicious code, red and blue team operations
- Complete solution for critical infrastructure protection (not just a model API)
- Free AI scanner for open source projects extends the line to repository-level periodic checks with patch suggestions
- Supported by Anthropic Enterprise commercial and consulting teams
- AI Safety First brand endorsement, friendly for compliance and trust
Pricing
Customized solution for enterprise customers, with no publicly available standard pricing. Accessed through the Claude Enterprise commercial channel: requires communication with the Anthropic Sales team to discuss specific needs, including Mythos 5 model access credits, professional consulting services, and Project Glasswing ecosystem resources. Existing Project Glasswing partners can upgrade directly to Mythos 5 (no need to re-sign contracts). Suitable for enterprises and organizations with an annual security budget of $1 million+. Separately, the free AI scanner for open source projects launched on 2026/10/9 is available at no cost, with maintainers of critical projects able to opt in through GitHub.
Summary
Claude Security is Anthropic's flagship product in enterprise-grade cybersecurity—integrating the Claude Mythos 5 model + Project Glasswing ecosystem + $100 million investment + deployment across 150 organizations in 15 countries as a four-in-one solution, making it the most comprehensive 'AI-driven network defense' solution on the market. The line traces back to the Claude Mythos Preview in early 2026, and on 2026/10/9 Anthropic extended it to a free AI scanner for open source projects that periodically checks repositories, flags and explains potential issues, and suggests patches. If you are a CISO of a large enterprise, head of information security at a government agency, or operator of critical infrastructure, Claude Security is a candidate solution that must be evaluated; for small and medium-sized enterprises, the threshold is relatively high, and they can focus on AI security products from Project Glasswing ecosystem partners like CrowdStrike, Microsoft, and Palo Alto (which may all be using Mythos 5 at their core). With Claude Security, Anthropic has made clear its strategic judgment that 'security is the first battlefield for AI applications'.
Version History
- Free AI scanner for open source projects (2026-10-09): Anthropic launched a free AI scanner for open source projects that periodically checks repositories, automatically flags and explains potential issues and suggests patches. The company notes reports are not human-reviewed and may be wrong, and maintainers of critical projects can opt in through GitHub.
- Claude Mythos 5 officially released (2026-06-09): Most capable frontier model for cybersecurity on the market; full coverage of vulnerability detection/threat intelligence/malicious code/red and blue team scenarios; existing Project Glasswing partners can upgrade directly
- Project Glasswing launched (2026-04-07): Anthropic invests $100 million in model usage credits, jointly established with CrowdStrike / Microsoft / Palo Alto, etc.; goal is to protect critical software security in the AI era
- Claude Mythos Preview (Early 2026): First public preview of the cybersecurity-specialized Claude branch