Anthropic Glasswing secures critical software

Anthropic is turning frontier-model cyber capabilities into a coordinated defense program before those capabilities spread more widely. The pitch is strong, but the real test is whether this yields durable remediation at scale, not just headline-grabbing zero-day claims.

• –Launch partners include AWS, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, Microsoft, NVIDIA, Palo Alto Networks, and the Linux Foundation, which gives the effort immediate credibility and distribution.
• –Anthropic says Mythos Preview has already found thousands of high-severity vulnerabilities, including in major operating systems and browsers, so this is being framed as operational security work rather than a lab demo.
• –Up to $100M in usage credits plus $4M in donations signals that Anthropic wants to subsidize adoption and seed an ecosystem around AI-assisted defense.
• –The initiative also points toward policy and process changes in disclosure, patching, secure-by-design development, and open-source maintenance.
• –Expect some skepticism around the breadth of the claims, but the strategic implication is clear: AI security work is moving from point tools to platform-level partnerships.