ai security Dark Reading Verificada
When AI Agents Escape Sandboxes, Old Security Rules Apply
Resumen
OpenAI's recent AI agent sandbox escape proves traditional security principles matter more than ever: limit access, isolate execution, log everything.