EU cybersecurity agency gains access to Anthropic Mythos

Anthropic is giving ENISA access to its Mythos model through Project Glasswing, a restricted program for defensive security testing.

Mythos can detect, combine, and exploit software vulnerabilities, and has already uncovered thousands of flaws, including zero-days.

The program lets participants share findings with external security and regulatory groups, broadening coordinated cyber defense.

The move expands defensive AI use in Europe, while underscoring dual-use concerns around advanced vulnerability-finding systems.