Anthropic has officially launched Project Glasswing, a collaborative initiative designed to leverage artificial intelligence for the defense of critical software infrastructure against AI-powered cyberattacks. The project brings together a coalition of over 45 organizations, including direct competitors such as Amazon Web Services, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, and Microsoft. At the core of this effort is Mythos, a new AI model currently in preview that Anthropic claims has identified security vulnerabilities across every major operating system and web browser tested.
The initiative represents a significant shift in how the technology sector approaches cybersecurity, moving from reactive patching to proactive vulnerability discovery. According to Anthropic, the Mythos model is capable of flagging system weaknesses with virtually no human intervention. Ars Technica reports that access to the Claude Mythos Preview is currently limited to a select group of high-profile customers, while TechCrunch confirms the model is being deployed for defensive cybersecurity work. This exclusivity contrasts with the broader public announcement, suggesting a phased rollout intended to manage the risks associated with such powerful autonomous security tools.
The project has drawn comparisons to historical scientific mobilizations, with ZDNet describing the alliance of twelve tech rivals as an "AI's Manhattan Project" for cybersecurity. This framing highlights the unprecedented nature of competitors collaborating to secure shared infrastructure against a common threat: advanced AI agents capable of exploiting software flaws. Wired notes that the collaboration aims to test and advance AI cybersecurity capabilities before adversaries can weaponize similar technologies. Gizmodo adds context by noting Anthropic's previous warnings that its newest model "presents unprecedented cybersecurity risks," underscoring the dual-use nature of the technology.
While the primary focus is on defense, the launch also touches on broader enterprise and cultural implications. Wired reports that Anthropic is simultaneously working to lower the barrier for businesses building AI agents with Claude, suggesting that robust security infrastructure is a prerequisite for wider agent adoption. The project's scope extends beyond commercial software; the involvement of government entities and critical infrastructure partners implies a national security dimension. By uniting rivals under a single defensive framework, Anthropic is attempting to standardize the detection of vulnerabilities that could otherwise be exploited in an increasingly automated threat landscape.
Market analysts and industry observers note that the success of Project Glasswing will depend on the model's ability to distinguish between benign code and malicious exploits without generating false positives. As the technology matures, the initiative may set a new precedent for how large-scale software security is managed in an era where AI development outpaces traditional human-led auditing processes.