Anthropic's Project Glasswing: How AI is Accelerating the Discovery of Thousands of Open-Source Vulnerabilities

The Ultimate Digital Detective

Imagine you have a magnifying glass that can look at a million pages of a book in one second and find every single spelling mistake. Now imagine that book is the code for the entire internet, and the spelling mistakes are security holes that hackers can use to steal information. For years, finding these holes has been a slow, boring job for human security experts. But in April 2026, the AI company Anthropic announced that they have built a magnifying glass so powerful it changes everything. They call it "Project Glasswing."

Project Glasswing is not just one company working alone; it is a massive team of tech giants. Anthropic brought together Amazon Web Services, Apple, Google, Microsoft, NVIDIA, Cisco, CrowdStrike, and the Linux Foundation. They are all working together to use Anthropic’s unreleased, super-secret AI model called "Claude Mythos Preview." This model is so good at reading code that it can find security vulnerabilities faster and more accurately than almost any human on Earth.

A $100 Million Promise to Open Source

Usually, when a company builds a super-powerful AI, they keep it hidden and sell it for millions of dollars. But Anthropic is doing something different. They realized that if this AI can find holes in software, hackers will eventually build their own AIs to find those same holes. The only way to win is to use the AI to fix the holes first. So, Anthropic is giving away up to $100 million in free AI credits to over 40 organizations that maintain critical open-source software. They are also donating $4 million directly to open-source security groups like the Open Source Security Foundation (OpenSSF) and the Apache Software Foundation.

This means that the volunteers who maintain the free software we all use can now use the world’s most advanced AI to scan their code for free. They can find the "loose boards" in their digital treehouse and fix them before the bad guys even know they are there. Project Glasswing is a perfect example of how AI can be used as a shield to protect the open-source world, rather than a sword to attack it. It is a new era where the biggest tech companies in the world are teaming up to make the internet safer for everyone.