What Is Project Glasswing?

What it is: Project Glasswing is Anthropic’s program that gives major tech companies (Apple, AWS, Google, Microsoft, Nvidia) early access to Claude Mythos for defensive security work.
Who it’s for: Anyone tracking AI security news in 2026.
Best if: You’re reading coverage of the May 2026 Apple/Mythos exploit and want to know what “Project Glasswing” actually refers to.
Skip if: You work on AI policy or enterprise security partnerships at one of the Glasswing partner companies.

Project Glasswing is Anthropic’s initiative to put its frontier model Claude Mythos in the hands of defensive security teams before equivalent capability is widely available to attackers.

The launch partners announced publicly are AWS, Apple, Google, Microsoft, and Nvidia — five of the largest pieces of consumer technology infrastructure. The premise: if frontier AI models can find vulnerabilities faster than ever, the people running the largest pieces of consumer software should have the model first, scan their own code with it, and ship patches before the same capability becomes generally available to bad actors.

The initiative gained widespread attention in May 2026 when Calif, a Vietnam-based security firm, used Mythos to identify two memory-corruption bugs in macOS that they chained into a working exploit against Apple’s Memory Integrity Enforcement on M5 silicon — in five days.

Why it matters

Glasswing is the first formal coordination program of its kind among AI labs and infrastructure vendors. It signals a recognition that the economics of vulnerability discovery have shifted — defenders need explicit time-to-patch advantage to keep up with AI-accelerated offensive research.

For the broader AI industry, Glasswing is also a hedge against regulatory pressure. If Anthropic can demonstrate that it’s coordinating with major incumbents on defensive security, the company has a stronger position when policymakers debate frontier-model access restrictions.

Where you’ll see it

• In our Claude Mythos / Apple M5 exploit breakdown.
• In our Claude Mythos guide.
• In Apple, AWS, Google, Microsoft, and Nvidia security blogs and partnership announcements.
• In coverage of the May 2026 AI security news cycle.

You May Also Like

• Claude Mythos: Anthropic’s new frontier-tier model
• Claude Mythos Cracked Apple’s M5 Security in 5 Days
• Apple’s AI Strategy in Mid-2026
• What is MIE (Memory Integrity Enforcement)?
• Anthropic: the company behind Claude