AWS and Anthropic push AI-powered cybersecurity forward with Project Glasswing and Claude Mythos Preview

AWS and Anthropic have joined Project Glasswing, a new defensive cybersecurity initiative built around Anthropic’s unreleased model, Claude Mythos Preview. Anthropic says the model can find and exploit software vulnerabilities at a very high level, so access is limited to selected organizations working to secure critical software before similar capabilities spread more widely.

Claude Mythos Preview is not a public Claude release. Anthropic says it is an invitation-only research preview for defensive cybersecurity work, and participants can access it through the Claude API, Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry. Anthropic also says it committed $100 million in model usage credits to support Project Glasswing participants during the preview period.

BEST SPRING 2026 DEALS

Editor's Choice

Private Internet Access

Access content across the globe at the highest speed rate.

70% of our readers choose Private Internet Access

70% of our readers choose ExpressVPN

ExpressVPN

Browse the web from multiple devices with industry-standard security protocols.

Nord VPN

Faster dedicated servers for specific actions (currently at summer discounts)

That makes the central idea in your draft real, but some of the extra AWS-specific claims in the sample do not appear in the official announcements I found. I found solid support for Project Glasswing, Mythos Preview availability through Amazon Bedrock, and AWS Security Agent reaching general availability. I did not find official support for the exact figures about 400 trillion network flows daily, 300 million malicious S3 encryption attempts in 2025, or the “6 hours to 7 minutes” review claim in the sources I checked, so I left those out.

Claude Mythos Preview is the centerpiece

Anthropic describes Claude Mythos Preview as a frontier model with unusually strong cybersecurity capabilities. In its Glasswing materials, Anthropic says the model can surpass almost all human experts at finding and exploiting software vulnerabilities, which is why the company chose a restricted rollout instead of a public launch.

Anthropic also says the model can help defenders uncover deep, long-standing flaws in critical software. Reuters reported that Anthropic launched the program with support from major companies including Amazon, Apple, Google, Microsoft, Nvidia, CrowdStrike, and Palo Alto Networks, with plans to expand access to additional organizations that manage critical software infrastructure.

The practical AWS angle is that Bedrock is one of the approved access paths. Anthropic’s official materials say Project Glasswing participants can use Mythos Preview on Amazon Bedrock, which gives AWS a direct role in how selected organizations test and deploy the model in enterprise environments.

AWS is also pushing its own AI security tooling

Separate from Anthropic’s model, AWS has also made AWS Security Agent generally available. AWS documentation describes it as a frontier agent that proactively secures applications across the development lifecycle with automated security reviews and context-aware penetration testing. AWS’s own weekly roundup also highlighted AWS Security Agent GA this week.

AWS positions Security Agent as a continuous testing and review system rather than a one-off scanner. Its documentation says it can validate security from design to deployment and deliver on-demand penetration testing across environments.

That means the AWS side of this story is twofold. First, AWS is one of the major partners in Anthropic’s Glasswing initiative. Second, AWS is expanding its own AI-driven security tooling with Security Agent as a generally available service.

What is official and what matters most

Why this partnership matters

This is one of the clearest signs yet that frontier AI companies and cloud providers expect cybersecurity to become a major battleground for advanced models. Anthropic’s position is that models like Mythos Preview can help defenders patch critical software before malicious actors gain the same level of automation. Reuters reported that Anthropic is also in discussions with the U.S. government as concern grows around AI-enabled cyberattacks.

For AWS customers, Bedrock’s inclusion matters because it gives selected enterprises a familiar cloud route for testing Mythos Preview under enterprise controls, instead of forcing them onto a separate experimental stack. Anthropic explicitly lists Amazon Bedrock among the supported access platforms.

The broader message is clear. Anthropic wants Mythos to help defenders first, and AWS wants to be one of the platforms through which that happens. At the same time, AWS is expanding its own AI-powered security tooling with Security Agent, which shows the company is not relying on partnerships alone.

Key points

• AWS is a named partner in Anthropic’s new Project Glasswing.
• Claude Mythos Preview is a restricted cybersecurity model, not a public Claude launch.
• Anthropic says Project Glasswing participants can use Mythos Preview through Amazon Bedrock.
• AWS Security Agent is now generally available as a separate AWS security product.

FAQ