Pwning OpenAI Atlas Through Exposed Browser Internals

Fascinating vulnerability. It shows how browser integrations can create unexpected attack surfaces. I’d love to see more discussion around hardening these kinds of local browser modules.