Anthropic launched “Project Glasswing,” a coalition that gives major technology firms and cybersecurity companies access to a preview version of its unreleased Claude Mythos model to strengthen defenses against AI-enabled cyberattacks. Anthropic says it is concerned that agentic coding capabilities could accelerate large-scale threats. The initiative provides participating organizations, including Amazon Web Services, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorgan Chase, Microsoft, and NVIDIA, the ability to use the Mythos Preview for defensive security work and to share learnings industrywide. Anthropic also made Mythos access available to roughly 40 additional organizations responsible for building and maintaining critical software infrastructure. Separately, Anthropic said it will no longer allow monthly Claude subscriptions to power third-party agentic tools, requiring API subscriptions and token-based fees for such use. The change is aimed at controlling distribution of subscription value while limiting misuse pathways. The developments underscore how university IT and research environments using frontier models may face heightened demands around secure deployment, governance, and documentation of AI-related risk management.