The Guardian’s take on Anthropic’s Claude Mythos: If AI finds every error, who controls the internet? | editorial

Anthropic announced its latest AI model, Claude Mythos, this month, but said it would not be released publicly because it turns computers into crime scenes. The company claimed it could find previously unknown “zero-day” vulnerabilities, exploit them, and essentially link these vulnerabilities to take over major operating systems and web browsers. Mythos did this autonomously, writing code and gaining privileges. The effects are significant. It’s as if a burglar was able to attack any building, get in, unlock any door, and empty any safe.

The Silicon Valley company has so far named 40 organizations as partners under Project Glasswing to help build a defense – and is asking them to “patch up” vulnerabilities before hackers have a chance to exploit them. All are Americans and sit at the heart of the U.S.-led digital system. Anthropic only shared Mythos with the UK outside the US, allowing the AI ​​Security Institute to test border models. After seeing it up close…