Anthropic built an AI model so powerful they decided they couldn't release it to the world.

Claude Mythos is a frontier model that was finding high-severity vulnerabilities in the world's most critical software. In every major operating system, in every major web browser. Thousands of them. It surpasses all but the most elite human security researchers at finding and exploiting flaws.

Anthropic essentially said: "this breaks everything if it gets into the wrong hands."

So they formed Project Glasswing, a consortium of AWS, Apple, Google, Microsoft, NVIDIA, Cisco, CrowdStrike, and JPMorgan to work together to secure the world's most critical infrastructure FIRST.

This is the cybersecurity equivalent of the Manhattan Project, except this time the scientists are choosing to lock it in a vault and only hand the keys to the people building the shields.

If your first reaction is "that's terrifying", good! Things as accelerating incredibly fast.

https://www.youtube.com/watch?v=INGOC6-LLv0
https://www.anthropic.com/glasswing