Anthropic launched Project Glasswing, an AI-driven cybersecurity initiative, to identify and mitigate critical software vulnerabilities using the unreleased Claude Mythos Preview model. This program provides controlled access to trusted partners to bolster defense systems against potential threats. The initiative was developed following internal testing where the model identified high-severity vulnerabilities across major operating systems, web browsers, and enterprise environments, including long-standing security flaws that remained previously undetected.

Rather than a public release, Anthropic maintains strict distribution, granting access only to pre-screened organizations to prevent the model's capabilities from being repurposed for offensive cyberattacks. The program currently involves approximately 200 organizations across more than 15 countries, including India, Japan, Germany, and South Korea. Participants span critical sectors such as banking and finance, healthcare, telecommunications, and energy.

The core of this program, Claude Mythos Preview, is an advanced AI model engineered for high-performance coding and reasoning tasks. Technical assessments indicate that the model can analyze massive codebases in seconds, detect zero-day vulnerabilities (previously unknown security flaws), and propose automated patches for high-severity issues. The system's performance is reported to be comparable to expert human security teams, having already discovered thousands of high-severity bugs across widely used software systems.

Project Glasswing includes a coalition of global technology and infrastructure companies, such as Microsoft, Google, Amazon Web Services (AWS), Apple, NVIDIA, Cisco, CrowdStrike, Palo Alto Networks, JPMorganChase, and the Linux Foundation. These organizations leverage the system for vulnerability detection, systematic code analysis, and infrastructure hardening to maintain robust digital security across global networks.

Anthropic launched Project Glasswing, an AI-driven cybersecurity initiative, to identify and mitigate critical software vulnerabilities using the unreleased Claude Mythos Preview model. This program provides controlled access to trusted partners to bolster defense systems against potential threats. The initiative was developed following internal testing where the model identified high-severity vulnerabilities across major operating systems, web browsers, and enterprise environments, including long-standing security flaws that remained previously undetected.

Rather than a public release, Anthropic maintains strict distribution, granting access only to pre-screened organizations to prevent the model's capabilities from being repurposed for offensive cyberattacks. The program currently involves approximately 200 organizations across more than 15 countries, including India, Japan, Germany, and South Korea. Participants span critical sectors such as banking and finance, healthcare, telecommunications, and energy.

The core of this program, Claude Mythos Preview, is an advanced AI model engineered for high-performance coding and reasoning tasks. Technical assessments indicate that the model can analyze massive codebases in seconds, detect zero-day vulnerabilities (previously unknown security flaws), and propose automated patches for high-severity issues. The system's performance is reported to be comparable to expert human security teams, having already discovered thousands of high-severity bugs across widely used software systems.

Project Glasswing includes a coalition of global technology and infrastructure companies, such as Microsoft, Google, Amazon Web Services (AWS), Apple, NVIDIA, Cisco, CrowdStrike, Palo Alto Networks, JPMorganChase, and the Linux Foundation. These organizations leverage the system for vulnerability detection, systematic code analysis, and infrastructure hardening to maintain robust digital security across global networks.