Native sandbox support now allows OpenAI agents to execute code and manage files within isolated environments. This update reduces the risk of agents damaging host systems during complex task execution. Developers can now deploy autonomous workflows with higher security guarantees. It is a practical, incremental improvement for agent reliability.