The Canadian federal government has gained access to Anthropic’s Claude Mythos AI, a defensive cybersecurity tool, through the invitation-only Project Glasswing preview program.
This partnership allows the Canadian Centre for Cybersecurity to identify previously undiscovered software vulnerabilities. By detecting these gaps early, the government aims to protect critical infrastructure, and national institutions from potential cyberattacks.
Minister Evan Solomon said the move is a very important step for Canadians and the government to ensure they can protect their institutions and the country. Solomon said the participation in Project Glasswing provides early access to the technology specifically to test for critical software vulnerabilities.
According to an Anthropic spokesperson, Claude Mythos is designed to discover vulnerabilities across every major operating system and web browser [3]. The tool focuses on defensive applications, allowing security professionals to find and patch holes before they can be exploited by malicious actors.
Anthropic is currently expanding the reach of the tool. The company is adding 150 new partners across more than 15 countries [1]. This expansion is part of a broader plan to grow the total number of Glasswing partners to approximately 200 [2].
The rollout began in May 2026, with further implementation details emerging this week. The Canadian government's inclusion in the program places it among a small group of global partners testing the AI's capability to secure digital environments.
“"We’re now part of Project Glasswing, which will give us early access to Anthropic’s Mythos AI to test for critical software vulnerabilities."”
Canada's adoption of Claude Mythos AI signals a shift toward proactive, AI-driven defense in national security. By integrating a tool capable of scanning all major operating systems, the government is attempting to move faster than adversaries who use similar AI techniques to find zero-day vulnerabilities. This partnership also aligns Canada with a growing coalition of about 200 global partners, standardizing defensive AI capabilities across multiple borders.
