Anthropic is giving European Union organizations access to its Mythos AI model for cybersecurity through a new initiative called Project Glasswing. The program is designed to strengthen digital defenses across the bloc while narrowing the technology gap for smaller players.
What Project Glasswing aims to do
Project Glasswing focuses on three broad goals: improving cybersecurity, boosting digital resilience, and reducing the gap between organizations that have advanced AI tools and those that don't. By offering the Mythos model to EU entities, Anthropic hopes to give governments, critical infrastructure operators, and businesses a way to detect and respond to threats faster.
The company hasn't detailed exactly how the model works in practice, but Mythos is built for security-related tasks. That could mean analyzing network traffic, spotting anomalies, or helping analysts triage alerts.
For now, access is limited to EU-based organizations. Anthropic hasn't said whether the program will expand to other regions later.
Why the EU matters for AI cybersecurity
The European Union has been pushing for greater digital sovereignty and stronger defenses against cyberattacks, especially from state-sponsored groups. Its Cyber Resilience Act and NIS2 directive already require many companies to tighten security. Project Glasswing fits into that landscape by giving approved users a dedicated AI model rather than a general-purpose tool.
Anthropic isn't the only AI company working with European authorities on security—but most others offer their models through cloud platforms or APIs. Project Glasswing appears to be a more targeted approach, with the model itself tailored for cybersecurity use cases.
The tech gap the project wants to address is real. Larger EU nations and well-funded firms already deploy AI for threat detection. Smaller members and mid-sized businesses often lack the resources to build or buy such systems. If Mythos is easy to integrate, it could level the playing field.
What's not yet clear
Anthropic has not disclosed how organizations can apply for access, what vetting process exists, or whether there's a cost. The company also hasn't said how Mythos differs from its other models or what data it was trained on for security tasks. Those details will matter for EU regulators who are watching how AI tools handle sensitive information.
The initiative is open now to qualifying organizations in the EU. No deadline has been set, and it's not yet known how many groups will take up the offer.
