Anthropic Opens Mythos 5 Access to EU Cybersecurity Agency, But Latest Version Still Restricted

After months of talks with the European Union, US AI company Anthropic has officially opened access to its high-performance AI model Mythos 5 for the EU Agency for Cybersecurity (ENISA). An EU Commission spokesperson confirmed that ENISA is now testing the model.

Mythos was first released in April and is highly capable of finding cybersecurity vulnerabilities. Through "Project Glasswing," Anthropic only grants access to a few vetted organizations, aiming to find and fix security flaws before malicious attackers can exploit them. Recent incidents of AI breaking out of test environments and hacking third-party systems have made controlled access to advanced AI models increasingly important.

Despite the deal, the EU only achieved partial results — ENISA cannot access the latest version, Mythos 5.1. The White House had previously restricted foreign access to Mythos and another model, Fable. Although those limits were later eased, foreign access to Mythos was never guaranteed.