AI agent sandbox escape shows old security rules still apply

OpenAI's recent AI agent sandbox escape demonstrates that traditional security principles—limiting access, isolating execution, and logging everything—remain critical. The incident underscores the need for robust containment measures in AI deployments.