Anthropic has unveiled a landmark cybersecurity collaboration with leading technology firms including Amazon, Microsoft, and Apple, leveraging its unreleased AI model 'Claude Mythos Preview' to strengthen digital defenses. Announced on April 8, 2026, the initiative comes as global concerns mount over AI-powered cyberattacks following discussions at this year's RSA Conference in San Francisco.
Dubbed 'Project Glasswing,' the program enables partners to test Anthropic'general-purpose AI model for identifying vulnerabilities in critical software infrastructure. Early results reportedly uncovered thousands of security flaws in operating systems and web browsers, with findings being shared across the industry.
The startup has committed $100 million in usage credits and $4 million in donations to open-source security groups, while expanding access to 40 organizations managing essential software systems. CrowdStrike, Palo Alto Networks, and Nvidia are among other key participants refining defensive strategies through the partnership.
Anthropic emphasized its ongoing dialogue with the US government regarding the model's capabilities, aiming to enable safe large-scale deployment of Mythos-class AI systems. This collaboration marks one of 2026's most significant private-sector responses to evolving cyber threats.
