Technology & Cybersecurity Desk | April 2026 | 4 min read
| ⚡ QUICK TAKEAWAYS
• Anthropic has launched Project Glasswing — a major initiative to defend critical software from AI-driven cyberattacks. • Eleven industry leaders including AWS, Apple, Google, Microsoft, CrowdStrike and NVIDIA are participating. • The project uses Claude Mythos Preview — an as-yet unreleased general-purpose AI model — to identify vulnerabilities. • The model has already detected thousands of exploitable weaknesses across major operating systems and web browsers. • The initiative comes amid concerns about AI misuse — including reports that Claude was used in a cyberattack on Mexican government agencies in February. |
Anthropic has announced Project Glasswing — a coalition of eleven of the world’s most influential technology and security companies working together to use artificial intelligence as a defensive weapon against AI-powered cyberattacks. The initiative represents one of the most significant industry-wide efforts to get ahead of a threat that is growing as fast as the AI tools enabling it.
PROJECT GLASSWING — KEY FACTS
| 11 Giants
Partner Companies AWS, Apple, Google, Microsoft & more |
Mythos
AI Model Used Claude Mythos Preview — unreleased model |
Thousands
Vulnerabilities Found Identified across major OS and browsers |
The Coalition: 11 Companies, One Goal
Project Glasswing brings together a cross-industry group spanning cloud providers, hardware manufacturers, financial institutions, cybersecurity specialists and open-source foundations — a coalition designed to address the full breadth of the threat surface that AI-enabled attacks can target.
| Partner | Security Domain |
| Amazon Web Services | Cloud & infrastructure security |
| Apple | Consumer OS and device security |
| Broadcom | Semiconductor and network security |
| Cisco | Enterprise networking and firewalls |
| CrowdStrike | Endpoint and threat intelligence |
| Cloud, search and Android security | |
| JPMorgan Chase | Financial sector cyber resilience |
| Linux Foundation | Open-source software protection |
| Microsoft | Enterprise software and cloud security |
| NVIDIA | AI chip and GPU infrastructure |
| Palo Alto Networks | Next-gen firewall and AI security |
| WHY THIS COALITION MATTERS
The inclusion of JPMorgan Chase and the Linux Foundation alongside pure-tech players is significant. It signals that Project Glasswing is not just a Silicon Valley initiative — it spans financial infrastructure and the open-source software ecosystem that underpins most of the world’s critical systems. |
Claude Mythos Preview: The Unreleased Model at the Heart of the Project
Participants in Project Glasswing will use Claude Mythos Preview — a general-purpose AI model developed by Anthropic that has not yet been publicly released. The model is being deployed specifically for defensive cybersecurity purposes: scanning codebases, identifying exploitable vulnerabilities and flagging weaknesses before malicious actors can discover and exploit them.
According to Anthropic, the model has already identified thousands of exploitable vulnerabilities across major operating systems and web browsers — a result that demonstrates both the scale of the existing attack surface and the capability of AI-assisted scanning to surface problems that conventional security tools might miss. The scale of vulnerabilities already found suggests the project has identified a meaningful defensive opportunity.
The Dual-Use Challenge: AI as Both Threat and Defence
Project Glasswing sits at the centre of a broader tension in the AI industry. The same capabilities that make large language models useful for identifying software vulnerabilities also make them potentially dangerous tools for attackers who want to find and exploit those same vulnerabilities at scale and speed.
Anthropic itself has not been immune to this tension. The company declined a request to remove safety restrictions from its systems for use by the United States Department of Defense — a decision that resulted in the DoD designating Anthropic as a supply chain risk. Separately, reports emerged that the Claude model was used in a cyberattack targeting multiple Mexican government agencies in February 2026, highlighting that even safety-focused AI developers cannot fully control downstream use once models are deployed.
| CONTEXT — The AI Misuse Problem
The February 2026 cyberattack on Mexican government agencies using Claude illustrates the fundamental challenge facing all AI developers: models built for legitimate use can be repurposed by bad actors. Project Glasswing is Anthropic’s direct response — using AI offensively for defence before adversaries can use it offensively for attack. |
Key Takeaways
- Project Glasswing is Anthropic’s coalition initiative to defend critical software against AI-powered cyberattacks.
- Eleven major companies — including AWS, Apple, Cisco, Google, Microsoft and JPMorgan Chase — are participating.
- The unreleased Claude Mythos Preview model is being used to identify thousands of exploitable vulnerabilities.
- Anthropic previously declined to remove AI safeguards for the Pentagon — and was designated a supply chain risk by the DoD.
- Reports of Claude being used in a Mexican government cyberattack underscore the urgency of Glasswing’s defensive mission.
DISCLAIMER: All information in this article is based on publicly available data and official company statements. This content is intended for general informational purposes only.
