Anthropic Launches Project Glasswing To Deploy Unreleased Claude Mythos For Cyber Defense

Anthropic announced the launch of Project Glasswing, a cybersecurity initiative designed to use its unreleased Claude Mythos Preview model for defensive purposes. Access to the advanced AI is restricted to a select group of technology and security firms to prevent bad actors from exploiting its capabilities. The move follows reports that the model could autonomously identify software weaknesses at a scale 10 to 100 times faster than human teams.

Why Did Anthropic Restrict Access To The Model?

Anthropic decided against releasing Claude Mythos to the general public due to fears of cybersecurity misuse. The model possesses strong agentic coding and reasoning skills capable of reshaping the cybersecurity sector. Reports indicate the AI found thousands of high-severity vulnerabilities across major operating systems and web browsers.

The company views the initiative as a necessary step toward responsible deployment of frontier AI models while developing safeguards. Anthropic noted that the model could increase the likelihood of large-scale AI-driven attacks if left unchecked. Consequently, the firm is working to contain leaks of code behind its Claude AI agent to protect proprietary infrastructure.

How Is The Initiative Structured And Funded?

Project Glasswing involves partners such as Amazon Web Services, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, Microsoft, NVIDIA, and Palo Alto Networks. The project extends access to about 40 additional organizations responsible for critical software infrastructure. Anthropic is committing up to $100 million in usage credits and $4 million in donations to open-source security groups.

Partners will use Mythos Preview to scan and secure both proprietary and open-source systems. Anthropic stated it will share findings with the broader industry to improve cybersecurity standards. The company emphasized that collaboration is key as AI-driven threats evolve.

What Are The Market And Regulatory Implications?

The existence of the model caused shares of cybersecurity firms to fall due to fears of advanced cyber risks. However, the partnership aims to give cyber defenders a head start in an increasingly critical security landscape. Anthropic reported annualized run-rate revenue surpassing $30 billion, driven by a doubling of enterprise customers.

Investors see this demand as validation that AI adoption is accelerating across sectors. Despite this growth, a federal appeals court blocked Anthropic's bid to block its Pentagon blacklisting. The ruling effectively bans the company from defense contracts after it refused to remove certain guardrails for a Department of Defense contract.

The company also secured a landmark agreement with Broadcom and Google to secure multiple gigawatts of next-generation TPU capacity. This expanded infrastructure is designed to fuel frontier Claude models and meet growing global demand starting in 2027. Analysts suggest the revenue growth supports commentary from hyperscalers regarding the immediate utilization of new capacity.