Anthropic launches Project Glasswing

2026-04-15 · anthropic

Anthropic is pushing frontier coding ability into defensive security with Project Glasswing. The initiative pairs Claude Mythos Preview with major partners to hunt vulnerabilities and harden critical software. The pitch is simple, but big, give defenders better agentic tooling before attackers get there first.

Key Features or Updates

Project Glasswing brings together AWS, Apple, Google, Microsoft, NVIDIA, and other partners around defensive scanning. Anthropic says Mythos Preview has already found thousands of high-severity vulnerabilities, and the company is backing the effort with up to $100M in usage credits plus $4M in donations.

Impact on Developers

For security teams, this is a signal that AI coding models are crossing into serious vuln research. Open-source maintainers and infra teams could get faster scanning and remediation help, but it also raises the bar for secure development practices.

How to use it

Use the announcement as a cue to review your own code scanning and dependency hygiene. If you work in security or platform engineering, watch for the follow-up reports Anthropic says it will publish from the pilot.

Read Original Post →