Anthropic Launches Project Glasswing With Major Tech Partners to Target Software Vulnerabilities

According to a recent LinkedIn post from Anthropic, the company is introducing Project Glasswing, an initiative aimed at securing critical software using its new frontier model, Claude Mythos Preview. The post indicates that Mythos Preview is designed to detect software vulnerabilities at a level comparable to highly skilled human experts and has already identified thousands of high-severity flaws across major operating systems and web browsers.

The company’s LinkedIn post highlights partnerships with major technology and financial players, including Amazon Web Services, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, Microsoft, NVIDIA, and Palo Alto Networks. These collaborations suggest Anthropic is positioning Mythos Preview as an infrastructure-level cybersecurity tool embedded in core systems rather than a broadly distributed product.

As shared in the post, Anthropic does not plan to make Mythos Preview generally available, instead focusing on controlled deployments and developing safeguards to block potentially dangerous outputs before wider rollout of Mythos-class models. The company plans to test these safeguards with an upcoming Claude Opus model, which may signal a gradual commercialization pathway that balances security concerns with monetization opportunities.

The post further notes a commitment of up to $100 million in Mythos Preview usage credits for partners and more than 40 additional organizations that maintain critical or open-source software. For investors, this credit-based model resembles a customer acquisition and ecosystem-building strategy that could accelerate adoption while delaying near-term revenue, potentially supporting long-term recurring demand if Mythos-based security capabilities prove differentiated.

Anthropic’s emphasis on collaborating with industry, open source projects, researchers, and governments suggests an effort to influence emerging standards for AI-assisted cybersecurity. Should Mythos Preview or its successors become foundational to vulnerability discovery in critical infrastructure, Anthropic could strengthen its strategic position relative to other AI model providers and deepen integration with large enterprise and cloud platforms.

However, the decision not to release Mythos Preview broadly and the focus on safety research indicate that near-term financial upside may be tempered by governance and risk-mitigation priorities. For the wider AI and cybersecurity sectors, the post implies that advanced models will likely play dual roles in offensive and defensive security, potentially increasing demand for trusted providers capable of safely operating high-capability systems.