Project Glasswing is a multi-organization cybersecurity initiative anchored by a single alarming capability: Claude Mythos Preview, an unreleased Anthropic frontier model, can find and exploit software vulnerabilities better than nearly every human security researcher alive.
The coalition includes AWS, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, Microsoft, NVIDIA, and Palo Alto Networks. That list is not a marketing slide. It reflects how seriously these organizations are taking the threat implied by Mythos Preview's coding performance, which Anthropic describes as surpassing all but the most elite human practitioners.
The full Anthropic writeup at anthropic.com/glasswing is worth reading for what it reveals about the internal evaluation process that triggered this coalition, not just the coalition itself. The model is unreleased. The vulnerabilities it can find are real. The question the piece forces is what happens when this capability is no longer controlled by one lab.
[WATCH ON YOUTUBE →]