SafetyExplainer
4 questions about reported OpenAI agent activity at 55 websites
The firm's public-data review cannot establish whether agents obtained sensitive information. A separate Australian incident shows why the boundary between public pages and other infrastructure matters.
Safety Desk