AI browsers are secure, promise.

The brilliant minds at the University of Washington have discovered that our futuristic agentic AI browsers are completely allergic to basic same-origin policies. This hilarious revelation leaves users of ChatGPT Atlas, Chrome with Gemini, Claude, or Perplexity Comet holding a rather empty security bag. To protect ourselves from embedded websites stealing our tab data, we are left to awkwardly avoid these vulnerable environments.