What is Project Glasswing?

The window between discovering a software vulnerability and an adversary exploiting it has officially collapsed—what used to take months now happens in mere minutes thanks to AI. In an era where attackers are moving at unprecedented speeds, defending the world's digital infrastructure requires a massive upgrade.

Enter Project Glasswing, a newly announced initiative by Anthropic designed to secure the world's most critical software for the AI era.

What is Project Glasswing? At its core, Project Glasswing arms the defenders of our digital infrastructure with early access to Anthropic's newest and most advanced frontier model, Claude Mythos Preview. Built as a general-purpose model, Mythos Preview is Anthropic's most capable AI yet for coding and agentic tasks. Because it can deeply understand and modify complex software, it is uniquely equipped to identify and fix security vulnerabilities at a scale and pace that was previously impossible. In fact, the model has already successfully identified thousands of zero-day vulnerabilities across critical infrastructure.

To make this happen, Anthropic hasn't just released a model; they've assembled an absolute powerhouse of a tech alliance. The initiative’s launch partners include heavyweights like Amazon Web Services (AWS), Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, Microsoft, NVIDIA, and Palo Alto Networks.

Early Testing Shows Massive Potential The early returns from these tech giants are incredibly promising:

• Microsoft tested Mythos Preview against their open-source security benchmark, CTI-REALM, and reported "substantial improvements" compared to previous AI models.

• AWS is already testing the model within its own security operations, applying it to critical codebases to strengthen code and harden the model itself.

• Palo Alto Networks reported that over just a few weeks of using Mythos Preview, the model identified complex vulnerabilities that prior-generation AI models missed entirely.

A Lifeline for Open Source Software Perhaps the most engaging part of Project Glasswing is its focus on the open-source community. Historically, elite cybersecurity has been a luxury reserved for massive corporations with deep pockets, leaving open-source maintainers—who build the foundational software the whole world relies on—to fend for themselves.

Project Glasswing aims to change this equation by turning AI into a "trusted sidekick for every maintainer". To back this up, Anthropic is committing up to $100 million in model usage credits and making $4 million in donations to open-source security organizations. Beyond the launch partners, access has also been extended to over 40 additional organizations that maintain critical software infrastructure.

The AI Arms Race is Here While Project Glasswing is a massive leap forward for cyber defense, it also serves as a stark warning. As Lee Klarich of Palo Alto Networks points out, AI's ability to find hidden vulnerabilities signals a dangerous shift where malicious actors will soon be able to develop exploits faster than ever before.

There is no going back. There will be more attacks, and they will be faster and more sophisticated. With Claude Mythos Preview now available to defenders via the Claude API, Amazon Bedrock, Google Cloud's Vertex AI, and Microsoft Foundry, the race is on. The defenders have a powerful new tool in their arsenal, but as the industry leaders agree: now is the time to aggressively modernize our cybersecurity stacks before the attackers beat us to it