Anthropic announced Project Glasswing, a critical-software security coalition built around its Claude Mythos Preview model. The launch partners are AWS, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorgan Chase, the Linux Foundation, Microsoft, NVIDIA, and Palo Alto Networks, with extended access for over 40 additional organizations that build or maintain critical software infrastructure. Anthropic is committing up to $100M in Mythos credits across the program plus $4M in direct donations to open-source security work. [1]

The paper's Tuesday piece on Anthropic's civilian door opening while the Pentagon door stays shut framed the institutional split. Wednesday's news fills in the corporate side and adds a quieter government detail: Anthropic has been in discussions with CISA and the Center for AI Standards and Innovation about Mythos's offensive and defensive capabilities, and the NSA is separately testing the same model — the model the Department of Defense has excluded from its Impact Level 6 and 7 network agreements signed with OpenAI, Google, Microsoft, SpaceX, and others. [2]

What makes the picture institutionally awkward is not that Anthropic has government interest. It is that the government interest is routing around the Pentagon. The NSA's parallel testing of a Pentagon-restricted model is a quiet institutional workaround. Glasswing is the corporate consortium. The civilian-agency engagement is in discussions, not contracts. The blacklist is holding at the DoD departmental level. It is not holding across the intelligence community. [2]

-- SAMUEL CRANE, Washington