CrowdStrike Joins Anthropic as Founding Member of Project Glasswing to Secure Claude Mythos Frontier AI

CrowdStrike joins Anthropic to secure the Claude Mythos model, providing sensor-level visibility and runtime governance for the next generation of frontier AI.

By: AXL Media

Published: Apr 8, 2026, 9:06 AM EDT

Source: Information for this report was sourced from CrowdStrike

Forging a Defensive Coalition for Frontier AI

CrowdStrike has officially announced its role as a founding member of Project Glasswing, a strategic coalition formed to secure the deployment of Anthropic’s latest frontier model, Claude Mythos. This partnership establishes a clear division of labor in the burgeoning AI industry: while Anthropic focuses on building high-capability models with native safety features, CrowdStrike provides the security infrastructure where the AI executes. The initiative aims to ensure that as AI becomes more autonomous and capable of complex planning, defenders maintain a definitive advantage over adversaries who are increasingly weaponizing the same technology.

Visibility as the Foundation of AI Governance

The deployment of the Mythos Preview introduces significant governance challenges as AI agents move beyond simple chat interfaces to accessing sensitive databases and financial systems. CrowdStrike’s contribution to the project centers on its ability to provide sensor-level visibility across trillions of events daily, allowing enterprises to discover every AI application in their environment. This level of insight is critical for managing "Shadow AI" tools deployed by employees without official security approval ensuring that no agent operates outside the oversight of corporate security policies.

Enforcing Runtime Protection for AI Agents

As AI agents begin to automate operations and reshape software development via tools like Claude Code, the risk shifts from model behavior to execution security. CrowdStrike’s AI Detection and Response (AIDR) capabilities are designed to secure all agents, whether homegrown or third-party, at the point of action. By monitoring what these agents access and what decisions they make in real-time, the Falcon platform provides a layer of machine-speed enforcement that prevents unauthorized workflows or the accidental exposure of proprietary enterprise data.