Anthropic has launched Project Glasswing, a new cybersecurity initiative bringing together Amazon Web Services, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, Microsoft, NVIDIA, and Palo Alto Networks to help secure critical software using advanced AI. At the center of the effort is Claude Mythos Preview, an unreleased frontier model that Anthropic said has demonstrated exceptional coding and vulnerability discovery capabilities, identifying thousands of serious software flaws across major operating systems, browsers, and other critical systems.
Anthropic said the initiative was created in response to rapid advances in AI model capability, which it believes are approaching a level where systems can outperform nearly all but the most elite human researchers in finding and exploiting vulnerabilities. Through Project Glasswing, the 12 launch partners, along with more than 40 additional infrastructure and software organizations, will use Mythos Preview for defensive security work across first-party and open-source systems. Anthropic said it would commit up to $100 million in usage credits and $4 million in direct support for open-source security groups.
At the same time, Anthropic announced expanded agreements with Google and Broadcom to increase compute capacity for its Claude models. The company said the new infrastructure, most of it based in the United States and coming online in 2027, would support surging demand, rising enterprise adoption, and the next phase of frontier AI development.
