When AI Agents Escape Sandboxes, Old Security Rules Apply

OpenAI's recent AI agent sandbox escape proves traditional security principles matter more than ever: limit access, isolate execution, log everything.