Introducing Project Glasswing
Anthropic has launched Project Glasswing, a significant undertaking dedicated to advancing artificial intelligence within the cybersecurity domain. At
the heart of this initiative is the Claude Mythos Preview, a sophisticated AI model designed by the San Francisco-based firm. This model stands out for its exceptional capabilities in handling coding and complex agentic tasks. Initially, its access is limited to select partners of Project Glasswing, empowering them to rigorously assess their technological frameworks for potential security weaknesses and vulnerabilities. The company has also committed substantial financial resources to fuel this crucial project, underscoring its importance in bolstering global digital defenses against escalating cyber threats.
Key Partnerships and Goals
The AI startup made a significant announcement on X, formerly Twitter, detailing the inauguration of Project Glasswing and its distinguished roster of launch partners. This elite group includes industry giants such as Amazon Web Services (AWS), Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, Microsoft, Nvidia, and Palo Alto. Anthropic articulates Project Glasswing as an "urgent initiative to help secure the world's most critical software." The Claude Mythos Preview, the exclusive model powering this endeavor, is made available not only to these premier partners but also to a broader network of over 40 organizations actively involved in the development or maintenance of essential software infrastructure. Furthermore, Anthropic has pledged substantial financial support, offering up to $100 million in usage credits and an additional $4 million in direct donations to open-source security organizations.
Claude Mythos Capabilities
Anthropic asserts that its specialized cybersecurity AI model, Claude Mythos, has already demonstrated remarkable efficacy by uncovering thousands of high-severity vulnerabilities. These discoveries span across every major operating system and widely used web browser, highlighting the model's comprehensive reach. The AI is reportedly adept at understanding and systematically altering intricate software code to pinpoint security flaws and subsequently devise solutions to rectify them. Partnering organizations are slated to integrate this powerful model into their ongoing defensive cybersecurity operations, enhancing their proactive threat detection and mitigation strategies.
Access and Pricing
Despite operating as a gated research preview, the Claude Mythos Preview is a commercial offering and is available as a paid model. The pricing structure is set at $25 per million input tokens and $125 per million output tokens. Access to this advanced AI is facilitated for its partners through various established platforms, including the Claude application programming interface (API), Amazon Bedrock, Google Cloud's Vertex AI, and Microsoft Foundry. Anthropic has explicitly stated that their intention is not to make Mythos Preview generally available to the public. Instead, their strategic objective is to deploy "Mythos-class models safely at scale," emphasizing the critical need for robust safeguards that can reliably prevent the model from generating its most dangerous outputs before broader deployment.
