Project Glasswing: Anthropic Unveils Global Alliance to Secure Critical Software Infrastructure
In an urgent bid to fortify the world's digital foundations, Anthropic has launched Project Glasswing—a massive, multi-partner initiative designed to preempt the next wave of AI-augmented cyberattacks.
Today, Anthropic announced the launch of Project Glasswing, a landmark security initiative bringing together a "who's who" of global technology. Partners including Amazon Web Services (AWS), Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, Microsoft, NVIDIA, and Palo Alto Networks have joined forces to secure critical software in the AI era.
The catalyst for this alliance is a breakthrough frontier model: Claude Mythos Preview. This unreleased system has demonstrated a leap in "agentic" capabilities, allowing it to autonomously discover and exploit software vulnerabilities that have survived decades of human and automated scrutiny.
The Mythos Capability: Defensive Urgency
During internal testing, Mythos Preview identified thousands of zero-day vulnerabilities in every major operating system and web browser. Its autonomy is what sets it apart, identifying flaws and chaining together exploits without human intervention.
Mythos's Security Track Record:
- OpenBSD (27-Year-Old Bug): Found a remote crash vulnerability in a system renowned for being "secure by default."
- FFmpeg (16-Year-Old Bug): Detected a critical flaw in code that had been hit 5 million times by automated testing tools without discovery.
- Linux Kernel: Autonomously found and chained vulnerabilities to escalate from a standard user to full system control.
"The window between a vulnerability being discovered and being exploited by an adversary has collapsed," warned Elia Zaitsev, Chief Technology Officer at CrowdStrike. "What once took months now happens in minutes with AI. That is a reason to move together, faster."
Bridging the Defense Gap
Anthropic is committing up to $100 million in usage credits for Mythos Preview to empower partners and over 40 additional organizations tasked with scanning and securing first-party and open-source systems. Furthermore, the company is donating $4 million to open-source security entities like the Linux Foundation and the Apache Software Foundation.
As part of the project, Cisco, Microsoft, and Google will integrate these capabilities into their defensive stacks. "Our foundational work with these models has shown we can identify and fix security vulnerabilities at a pace and scale previously impossible," said Anthony Grieco, SVP & Chief Security & Trust Officer at Cisco.
Benchmarks: A New Frontier in Agentic Coding
Anthropic's system card reveal shows Mythos Preview significantly outperforming previous models in specialized coding and reasoning benchmarks:
| Evaluation | Mythos Preview | Claude Opus 4.6 |
|---|---|---|
| SWE-bench Verified | 93.9% | 80.8% |
| SWE-bench Pro | 77.8% | 53.4% |
| CyberGym (Vulnerability Repro) | 83.1% | 66.6% |
The Impact for MENA Infrastructure
For the Middle East, particularly the UAE and Saudi Arabia, which are rapidly deploying large-scale digital infrastructure and "sovereign" AI clouds, Project Glasswing represents a critical shift in the threat model. The regional focus on smart cities, digital governance, and energy grids makes the proactive patching of common software like the Linux kernel and FFmpeg a top priority.
Anthropic intends to report publicly on the findings and vulnerabilities fixed within the next 90 days. The eventual goal is to enable wider deployment of Mythos-class models at scale, once robust cybersecurity safeguards are refined and battle-tested.
"Project Glasswing is a starting point," the announcement concludes. "The work of defending the world’s cyber infrastructure might take years; frontier AI capabilities are likely to advance substantially over just the next few months."