Following leaked information in late March, **Anthropic** officially announced Mythos Preview on Tuesday, accompanied by the formation of **Project Glasswing**. This consortium aims to tackle the cybersecurity implications of the new **Claude** model and the broader advancements in AI capabilities. ### Project Glasswing: An Industry Collaboration The group includes major players such as **Microsoft**, **Apple**, **Google**, **Amazon Web Services (AWS)**, the **Linux Foundation**, **Cisco**, **Nvidia**, **Broadcom**, and over 40 other organizations spanning tech, cybersecurity, critical infrastructure, and finance. These members will have exclusive access to the model, which is not yet publicly available. The primary goal is to allow these foundational tech platform developers to assess Mythos Preview within their own systems, identifying and mitigating potential vulnerabilities and exploit chains. More broadly, **Anthropic** emphasizes that **Project Glasswing** is intended to stimulate urgent exploration of how AI capabilities are poised to revolutionize software security and digital defense practices worldwide. ### A Paradigm Shift in Security According to Logan Graham, **Anthropic's** frontier red team lead, the implications extend far beyond a single model or company. "We need to prepare now for a world where these capabilities are broadly available in 6, 12, 24 months. Many things would be different about security. Many of the assumptions that we’ve built the modern security paradigms on might break." AI models are increasingly capable of identifying code vulnerabilities and suggesting mitigations, or even strategies for exploitation. This creates a new iteration of the classic security cat-and-mouse game, where tools can aid defenders but also empower malicious actors, making previously impractical attacks more feasible. ### Mythos Preview: A Significant Leap **Anthropic** CEO Dario Amodei highlighted that Mythos Preview represents a significant advancement. "We haven't trained it specifically to be good at cyber. We trained it to be good at code, but as a side effect of being good at code, it's also good at cyber." He emphasized the need for a proactive response to the increasing power of AI models. Graham noted that Mythos Preview can perform advanced exploit development, penetration testing, endpoint security assessment, system misconfiguration detection, and software binary evaluation without source code access. ### Coordinated Vulnerability Disclosure **Anthropic** is adopting a staggered release strategy for Mythos Preview, starting with industry collaboration, to leverage the principles of coordinated vulnerability disclosure. This approach gives developers time to patch vulnerabilities before public disclosure. "We've seen Mythos Preview accomplish things that a senior security researcher would be able to accomplish," Graham stated, highlighting the potential impact on how such capabilities are released to prevent attackers from gaining an advantage. ### Industry Enthusiasm and Collaboration **Project Glasswing** partners have expressed enthusiasm for this collaborative effort. Heather Adkins, **Google's** vice president of security engineering, stated, "Google is pleased to see this cross-industry cybersecurity initiative coming together. We have long believed that AI poses new challenges and opens new opportunities in cyber defense." Companies maintaining internet infrastructure and developing foundational tech platforms are particularly supportive, especially given that Mythos Preview has already uncovered thousands of critical vulnerabilities, including decades-old bugs that have been repeatedly overlooked. Igor Tsyganskiy, **Microsoft's** global CISO, noted, "As we enter a phase where cybersecurity is no longer bound by purely human capacity, the opportunity to use AI responsibly to improve security and reduce risk at scale is unprecedented. Joining Project Glasswing, with access to Claude Mythos Preview, allows us to identify and mitigate risk early and augment our security and development solutions so we can better protect customers and Microsoft." ### The Need for Global Collaboration **Anthropic's** Graham emphasized the urgency and necessity of global collaboration. "Probably the most important thing the group needs to do is figure out all the questions that need answers and then figure out the answers," Graham said. "Project Glasswing is the starting point. It will fail if it’s just a handful of companies using a model. It has to grow into something even larger."