Anthropic Eases Secrecy Rules Around Mythos AI Cybersecurity Program, Allowing Partners to Share Threat Intelligence

Anthropic relaxed secrecy rules for its Mythos cybersecurity program, expanding what approved partners can share.

Participants in Project Glasswing can now disclose findings, tools, and best practices to outside organizations, regulators, and government agencies under responsible-disclosure rules.

The shift is meant to speed collaboration on AI-driven cyber defense and threat intelligence.

It also highlights the dual-use challenge of frontier AI models that can strengthen defenses but also aid attacks.