Anthropic announced its new Claude Mythos Preview model and Project Glasswing, a consortium involving Apple, Google, and more than 45 other organizations. The initiative aims to test advancing AI cybersecurity capabilities amid growing concerns over powerful models. The formal reveal followed leaked details about the model at the end of March.
Anthropic formally unveiled the Claude Mythos Preview model on Tuesday, April 7, 2026. The company simultaneously introduced Project Glasswing, an industry-wide effort to address cybersecurity risks from advanced AI systems like this new model and others in development across the field. Anthropic convened the consortium to collaboratively explore these challenges using the Mythos Preview model for testing purposes. The project brings together major players including Apple and Google, alongside more than 45 other organizations. Participants will focus on evaluating vulnerabilities such as hacking, malware, and other security threats posed by increasingly capable AI. Keywords associated with the announcement highlight concerns over cybersecurity, security, hacking, hacks, malware, and vulnerabilities. This move comes after revelations leaked at the end of March about Anthropic's development of a powerful new Claude model. By partnering with rivals, Anthropic seeks to strengthen defenses against potential AI-driven exploits in a rapidly evolving landscape.
