Anthropic Teams with Rivals on AI Cybersecurity Project

In a landmark move for AI safety, Anthropic has announced it is spearheading Project Glasswing, a major coalition that includes tech titans Apple and Google alongside more than 45 other organizations. The initiative is focused on rigorously testing advanced AI cybersecurity capabilities before they become widely deployed. The central tool for this effort will be Anthropic's new Claude Mythos Preview model, which the consortium will use to probe the frontiers of AI security. The project has two primary, interconnected goals: first, to prevent powerful AI systems from being exploited for malicious hacking activities, and second, to harden critical infrastructure—such as power grids, financial networks, and communication systems—against next-generation AI-powered threats. This collaborative approach marks a significant shift in the industry, moving beyond isolated corporate efforts toward a shared, pre-emptive defense strategy. By bringing together leading AI developers, cybersecurity experts, and infrastructure operators, Project Glasswing aims to establish robust safety standards and red-teaming practices that could define how AI is secured for years to come. The formation of this coalition underscores the growing consensus that the security implications of advanced AI are too vast and complex for any single entity to manage alone.