News
White Paper
Why Enterprise AI Fails at Scale
Key Takeaways
- Anthropic's Claude Mythos Preview has already identified thousands of high-severity flaws.
- The 12 launch partners include AWS, Apple, Google, Microsoft, NVIDIA and JPMorganChase, with 40+ additional organizations also receiving access.
- Anthropic is backing the initiative with up to $100 million in usage credits and $4 million in donations to open-source security groups.
Anthropic believes its new powerful AI model can improve enterprise cybersecurity — if they can keep it under control.
According to Anthropic, the model — Claude Mythos Preview — has already identified thousands of high-severity vulnerabilities, including flaws in every major operating system and web browser.
The company committed up to $100 million in usage credits for Mythos Preview and $4 million in donations to open-source security organizations.
"AI capabilities have crossed a threshold that fundamentally changes the urgency required to protect critical infrastructure from cyber threats, and there is no going back."
- Anthony Grieco
SVP & Chief Security & Trust Officer, Cisco
Table of Contents
What Is Project Glasswing?
Named after the glasswing butterfly, Project Glasswing is Anthropic’s new cybersecurity initiative that centers on deploying Anthropic's unreleased Claude Mythos Preview model to find and fix vulnerabilities in widely used software.
The initiative brings together 12 initial launch partners:
- Amazon Web Services
- Anthropic
- Apple
- Broadcom
- Cisco
- CrowdStrike
- JPMorganChase
- The Linux Foundation
- Microsoft
- NVIDIA
- Palo Alto Networks.
Beyond the launch partners, Anthropic extended access to more than 40 additional organizations that build or maintain software infrastructure. The company said it plans to publish findings within 90 days.
How Does Claude Mythos Preview Work?
The initiative isn't a single tool — it's a suite of defensive capabilities that range from scanning source code to simulating real-world attacks. Here's how Anthropic says partners will put the model to work.
| Use Case | How It Works |
|---|---|
| Local vulnerability detection | Scans source code and systems for security flaws |
| Black box testing | Tests compiled binaries without access to source code |
| Penetration testing | Simulates attacks to find exploitable vulnerabilities |
| Open-source scanning | Scans open-source dependencies for known and novel flaws |
‘We Need to Act Now,’ Says Anthropic
"We formed Project Glasswing because of capabilities we’ve observed in a new frontier model trained by Anthropic that we believe could reshape cybersecurity."
- Anthropic Officials
The software we rely on everyday — think banking systems and logistics networks — has always had bugs, according to Anthropic. Some are minor. Others are major flaws that could allow cyberattackers to hijack systems, disrupt operations and steal data.
The financial consequences are already high. Research shows total global financial costs of cybercrime sit around an estimated $500 billion annually.
Project Glasswing is merely a starting point, according to Anthropic, adding that no single organization can solve these cybersecurity problems alone.
“Frontier AI developers, other software companies, security researchers, open-source maintainers and governments across the world all have essential roles to play,” according to an official statement.
To come out ahead, the company warned, “we need to act now.”