Anthropic has moved its newest frontier model into a restricted cyber-defense program rather than a public launch, betting that the safest first use for Claude Mythos Preview is inside a coordinated effort to harden critical software. In its announcement of Project Glasswing, the company said the initiative would give selected organizations early access to the model for defensive security work.

The program brings together launch partners including Amazon Web Services, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, Microsoft, Nvidia and Palo Alto Networks. TechCrunch reported that Anthropic is also extending the preview to more than 40 additional organizations and says the model has already identified thousands of serious vulnerabilities, including flaws in major operating systems and browsers.

Anthropic is framing the project as a race to put advanced AI systems in defenders' hands before similar capabilities spread more widely. VentureBeat reported that the company considers Mythos powerful enough to withhold from general release for now, even as it argues that security teams need early experience with the model to prepare for faster and more capable AI-assisted attacks.

That tension sits at the center of the rollout. Fortune reported that Anthropic is giving major technology and cybersecurity firms access to the preview specifically to strengthen defenses across critical systems, while the company says it will share lessons from the program more broadly. Anthropic is presenting Project Glasswing as a restricted security initiative rather than a broad commercial launch, with participating companies using the model to identify weaknesses before similar capabilities become more widely available.