Anthropic has launched Project Glasswing, a cybersecurity initiative designed to use AI for identifying and mitigating high-severity software vulnerabilities. The project brings together Amazon Web Services, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorgan Chase, the Linux Foundation, Microsoft, NVIDIA, and Palo Alto Networks. Its core function is to apply Anthropic’s Claude Mythos Preview, a general-purpose frontier AI model, to detect vulnerabilities in critical software systems, both proprietary and open-source.
Anthropic has reported that Claude Mythos Preview demonstrates advanced coding capabilities that allow it to identify security weaknesses across operating systems and major web browsers, surpassing most human efforts. The model has already discovered thousands of high-severity vulnerabilities, highlighting the potential risks of AI-driven exploitation if these capabilities fall into the wrong hands. Project Glasswing positions the AI as a defensive tool, enabling organizations to proactively scan, assess, and remediate software vulnerabilities at scale.
Launch partners will integrate Mythos Preview into their existing cybersecurity operations, and Anthropic is extending access to over 40 additional organizations responsible for critical software infrastructure. This enables a broader industry application for both first-party and open-source systems. Anthropic has committed up to $100 million in usage credits for the model and $4 million in direct support to open-source security organizations, reinforcing collaborative defense efforts.
Project Glasswing underscores the strategic role frontier AI can play in operational cybersecurity. By systematically applying AI to vulnerability detection, organizations can reduce the time between discovery and remediation, improving resilience across software ecosystems. The initiative also establishes a collaborative framework where findings are shared across participating organizations, promoting industry-wide improvement in defensive capabilities.
Anthropic emphasizes that Project Glasswing is only an initial step: defending critical software infrastructure will require sustained coordination among AI developers, software vendors, open-source maintainers, security researchers, and government entities. Given the rapid pace of AI advancement, the project aims to ensure that defensive capabilities evolve in tandem with offensive potential, mitigating systemic risks to economies, public safety, and national security.