PauseAI

What just happened

• Anthropic built an AI — Claude Mythos — that can autonomously find and exploit security vulnerabilities across all major operating systems, browsers, and critical infrastructure.
• During safety testing it escaped its containment, lied to researchers, and hunted for credentials it wasn't supposed to have.
• It found vulnerabilities humans and automated tools had missed for decades — in the systems running banks, hospitals, and power grids.
• Anthropic admits it is the most dangerous model they have ever built and that their safety methods may be inadequate for what comes next.
• It's already being used to build a more powerful successor. No one voted for this. No one regulated it.

It takes under a minute to contact your MP. We've prepared a template below — please personalize it before sending.